Milton Wexler picked up his daughters, Alice and Nancy, from the airport and drove them to his apartment. After an anxious ride, he sat them down and told them a story. The previous summer, a policeman had stopped their mother as she was walking across the street on her way to jury duty. “How can you be drunk so early in the morning? Shame on you!”, he shouted.

Leonore Wexler, a smart, law-abiding, 53-year-old retired biology teacher, wasn’t drunk. In fact, she rarely touched alcohol, let alone at nine in the morning. But she could see why the policeman thought so; she, too, had noticed her increasingly lumbering movement. Knowing what this might mean, she called Milton, who arranged an appointment with a neurologist that very same afternoon.

Leonore’s worst fear came true: she had Huntington’s disease.

The diagnosis was a nightmare she knew too well. Huntington’s had taken the lives of her father and her three brothers. Abraham Sabin, Leonore’s father, died in a state hospital on Long Island when she was 13. She overheard the doctor say he had Huntington’s chorea. After looking it up at the local library, she learned about the fatal neurological disease haunting her family. Chorea, from the Greek word for “dance,” cruelly expresses one of Huntington’s distinct symptoms: an uncontrollable writhing of the body, arms tracing arcs in the air, legs jerking in random directions, face twitching through a series of expressions — the opposite of dancing. She learned that there was no treatment and that, incorrectly, the disease afflicted only men.

Leonore went on to be the only one among her siblings to attend college. Keen to learn about her family’s affliction, she got her master’s degree in biology with a specialty in genetics at Columbia1. Her brothers took different paths: Jesse sold clothing, while Paul and Seymour started a swing band in New York City.

By 1950, the signs became impossible to ignore. Jesse, then 48, loved performing magic tricks, spinning coins, and pulling them out of his ears, nose, and pockets. Now, his fingers danced uncontrollably, and his coins fell to the floor. Paul, 44, and Seymour, 43, increasingly felt off-balance while walking and struggled to find words during conversations. In September 1950, a New York neurologist diagnosed all three brothers with Huntington’s on the same day.

Leonore was devastated by the news and descended into depression. Her brothers had funded her education, a path she took to understand the very disease now confirmed to haunt them. At 36, she was “fearful for what lay ahead for them and for herself.”

Milton hadn’t known about the disease in his wife’s family. When he learned that both men and women could be affected, he realized what this meant: Leonore might have passed it on to their daughters. It only deepened her anguish.

Facing his family’s genetic roulette, Milton, a former Navy lieutenant commander, sprang into action. He left the Menninger Foundation in Kansas in 1951 and moved his family to Los Angeles to start a more lucrative private practice, knowing he’d need to support his brothers-in-law and prepare for what might come. Despite the chaos, he clung to hope for his daughters. After Paul died in 1964, the second of Leonore’s brothers to pass, Milton assured Nancy and Alice that their mother would not get the disease. He truly believed that his daughters would be spared as well.

Subscribe now

Back in the apartment in 1968, after finally revealing to Alice and Nancy the family’s affliction, Milton explained that they had a 50-50 chance of getting the disease. If they got it, then their kids would have a 50-50 chance of getting it too. There was no test to say if they had it or not. The two sisters finally learned the truth about their family that had been hushed away since they were children: their grandfather and all three uncles died from Huntington’s, and now their mother was destined to repeat this fate.

That afternoon, Nancy and Alice hung onto each other and sobbed uncontrollably, terrified of the “grim roulette” their lives had just turned into. Either to ease their father’s pain or their own, they told him that a 50-50 risk wasn’t so bad. They later revealed that they remember very little of the conversation — just that their mother was dying, and that they decided not to have kids.

The sisters juggled their university education with watching their mother slowly deteriorate. “She was sad, silent, listless, vague. It was as if some dark subterranean river was taking her away from me. In retrospect, I do not know if her decline was psychological, neurological, or both. Perhaps the ominous gene was already beginning to hold,” Nancy would recount.

If the “dark subterranean river” was, in fact, an effect of the “ominous gene,” then it was a mere foreshock before an imminent earthquake. About a year after her diagnosis, in 1970, one night at 3 AM, Leonore’s in-home nurse found her sleeping deeply with an empty bottle of pills next to her. She called Milton, who called an ambulance. A shot of stimulants to the heart and a pump of the stomach at the hospital revived her, but it was a close call.

For Nancy, her mother’s suicide attempt marked a pivotal moment. She immediately shifted gears from concerned daughter to cure hunter, barrelling towards finding out everything about Huntington’s. The curse became a challenge.

While pursuing her PhD in clinical psychology at the University of Michigan in Ann Arbor, Nancy learned about a group of people with Huntington’s that met up in Detroit. In the evenings and weekends, she would drive the sixty-mile journey into the city to talk with them about their problems. The group eventually formalized into the Michigan chapter of the Committee to Combat Huntington’s Disease (CCHD), the organization founded by Marjorie Guthrie after her ex-husband, the famous folk singer Woody Guthrie, died from the disease. These conversations often left Nancy “feeling so depressed and exhausted that she would almost fall asleep during the long drive home.” Still, Nancy would quickly become the vice-president, then president of the chapter.

As Nancy immersed herself in the world of Huntington’s — university research, doctoral dissertation, CCHD meetings — Milton and Alice grew increasingly worried. One of the small tragedies within families with hereditary diseases like this is that every conversation and interaction begins to revolve around it, so much so that other events and dreams get drowned out. But barring occasional outbursts — “ENOUGH HD!!!!!!!!”, she wrote in a letter in 1973 — Nancy felt inspired rather than cursed by the resilience she witnessed among the interviewees, determined to “turn [herself] inside out to cure this thing.”

As her fame grew, another personal challenge loomed at home: her mother’s deteriorating condition. Leonore was “terrified, helpless, depressed and often over-medicated.” As her disease progressed, she was moved into a succession of nursing homes. Nancy would harrowingly chronicle her mother’s decline there:

She said she felt like she was in quicksand, trying desperately to keep from going under.

If she sat relaxed, her fingers kept up a constant motion, as if she were playing a sad tune on a silent piano. Her face twisted, her toes jumped. […] When she walked, her left side sagged, and her legs sometimes buckled suddenly, as if she had been hit at the back of her knees.

Nancy and Alice would eventually take on the role of caregiver:

As she became increasingly ill, I dressed her, carried her, helped her brush her teeth and go to the bathroom, fed her and, mostly, held her and kissed her. Her eyes still haunt me with their sadness and fear. Even possessed by chaotic violent movements, she could be graceful. Until close to the end, she had a sense of humor, and we could sometimes tease her from her worries. She always knew us.

Over time, her mother got frailer, her speech less intelligible, her twitching more chaotic. After a decade of living in nursing homes, on May 14, 1978, Leonore Wexler died.

Following the cremation, Nancy and Alice held a private service, reading letters from their mother’s early days, re-creating the vibrant, cheerful, intelligent woman they wanted to remember. They found themselves “strangely at peace, closer to our mother than we had in a long time.”

No Slides, No Speeches: The Hollywood Psychotherapist’s Workshop

In 1968, shortly after his wife’s diagnosis, Milton Wexler founded the Hereditary Disease Foundation (HDF). Its mission was simple in theory: to find the cause and cure for Huntington’s disease. Unlike the CCHD, which was more focused on patient care and education, the HDF was created to fund transformative research to find treatments and ultimately a cure.

Milton Wexler was in a favorable position to start this. Described as the “psychotherapist to the stars”, his practice counted actors, directors, singers, and architects among its clients — ideal for fundraising. The techniques he developed to bring out the best in creatives, particularly his “group process” encouraging bold, wide-ranging discussions, had proved to be successful with his artistic clients. Would they also work with scientists?

Milton thought so, and began organizing workshops unique for science at the time. They were multidisciplinary — bringing together scientists, clinicians, patients, and public figures — and they were small — just “fourteen to fifteen around a table, at most seventeen.” He recruited the brightest young minds regardless of domain, those familiar with the newest technologies, and paired them with experienced geneticists. There were to be no slides and no presentations, just an “open and unconstrained atmosphere” where everyone was encouraged to share their crazy ideas. Every workshop began with a family affected by Huntington’s sharing their story, reminding scientists, many of whom had never seen a person with the disease, of the human stakes of their research. They also came with some unique perks: Milton occasionally tapped into his Hollywood connections and, some nights, scientists might find themselves at a party in the home of a movie star.

It was at one of these workshops, in Bethesda in 1979, where disparate pieces of the Huntington’s puzzle would finally fall into place, revealing to Nancy a promising new avenue in her quest to understand the disease. For years, finding the Huntington’s gene had seemed like searching for a single sentence in a library of three billion letters. But new molecular tools were about to make that search feasible.

“Leaping gazelle-like through the genome”

Scientists have been trying to understand DNA, the code of life, for centuries. By the mid-1970s, thousands of illnesses had been identified as genetic, but the actual genes that caused the illness — often just one — were still unknown. If these diseases were to be cured, scientists sensed they first needed to be understood.

But with what tool? Scientists could learn how Huntington’s destroys families across generations, could document every cruel symptom, yet had no way to find the gene responsible. The question haunting researchers was simple but seemed impossible: how do you locate a single defective gene among three billion base pairs of DNA? What’s the difference between those with Huntington’s and those without?

By the late 1970s, many approaches to understanding Huntington’s had stalled. Investigating dopamine imbalances like those found in Parkinson’s, neurotransmitter deficiencies, and skin cell membrane defects had each led to negative findings, practical barriers, or inconsistent results.

But a new technology was emerging. In the early-70s, recombinant DNA techniques were revolutionizing molecular biology. Scientists had discovered that restriction enzymes, originally a defence mechanism of bacteria against foreign DNA, could work as molecular scissors that could recognize and cut DNA at specific sequences with exceptional precision. This ability made it possible to then clone fragments of human DNA and to visualize them through a technique called Southern blotting, revealing specific sequences as dark bands on film2.

Experimenting with these tools led to a crucial discovery. In the mid-1970s, David Botstein and Ronald Davis, geneticists from MIT and Stanford, found that humans carry small variations in DNA — sometimes just a one-base-pair difference — that usually did not cause any change in appearance or function. These harmless variations, called polymorphisms, were scattered throughout the genome.

When cut with a specific restriction enzyme, these variations changed where in the DNA the cut occurred, producing fragments of different lengths3. On Southern blots, these then appeared as distinct patterns, with the shorter fragments represented by bands near the bottom, and longer fragments closer to the top, creating a sort of molecular fingerprint. Because restriction enzymes revealed these polymorphisms through fragment length differences, they were called restriction fragment length polymorphisms, or RFLPs.

The breakthrough came when researchers realized RFLPs could serve as genetic landmarks. The principle, called linkage, is elegant: genes located physically close together on a chromosome tend to be inherited together. Consider a thought experiment: imagine the hemochromatosis gene sits on chromosome seven, and the gene governing hair texture is its immediate neighbor. If the defective hemochromatosis gene arose in an ancestor with curly hair, both variants travel together through generations — chromosomes rarely splinter. Over multiple generations, a statistical pattern emerges: curly-haired children in this family tend to have hemochromatosis4. By following the pattern of molecular fingerprints through a family tree, researchers could follow the disease — even without knowing what the disease gene looked like.

By 1978, these advancements in molecular biology had not yet been considered in the quest for understanding Huntington’s disease. That same year, at a conference in Utah, Botstein, Davis, Raymond White, Mark Skolnick, and other researchers had proposed using RFLPs to map the human genome. They theorized that roughly 150 RFLPs, spread like landmarks across all the chromosomes, could locate any gene through its inheritance pattern. The approach’s potential was demonstrated that year when researchers found an RFLP next to the beta-globin gene responsible for sickle-cell anemia, the first disease diagnosable prenatally using these markers. The culprit gene had been narrowed from an entire city to a single neighborhood.

Could it work for Huntington’s? Unlike sickle-cell anemia, where the faulty protein was known, Huntington’s remained mysterious. But the approach was promising. The central question became: how much work would it take to find an RFLP marker linked to the Huntington’s gene?

This was the focus of that pivotal HDF workshop in Bethesda in October 1979. Botstein, White, and several pioneering geneticists and molecular biologists gathered. As Nancy recalled, the meeting quickly turned into “total pandemonium” with everyone “yelling and screaming… and scribbling furiously.” Botstein and White felt that before looking specifically for the Huntington gene marker, a “map” of the whole genome was required, which they estimated would take about ten years. David Housman, the principal investigator at MIT, countered that that was too long a wait to start, and that they should test each newly identified RFLP as it emerged. It was a gamble — “leaping gazelle-like through the genome,” as Botstein put it, was an “enormously complex, time-consuming process” — but it could mean saving years.

Nancy was willing to gamble. Her reasoning was that even if it’s more work, testing the markers as they came along would increase the probability of finding the disease sooner; so they should do it. There was a buzz around this new technology, and everyone felt that it would eventually work. The only real issue was finding the right families. The technique required large kindreds spanning generations, ideally living close together under similar conditions. American families were not large enough and too scattered. “The real key to human gene mapping,” Botstein had realized, “was not finding the gene, but finding the humans.”

Nancy knew exactly where she might find such large families. Seven years earlier, at the 1972 Centennial Symposium in Columbus, Ohio, she had witnessed a remarkable presentation by Ramón Avila-Girón, who played a short film showing families afflicted by Huntington’s in the stilt villages surrounding Lake Maracaibo in Venezuela. The movie had left an indelible impression on her, and the idea that these families might play an important role in her quest was planted in her mind. In July 1979, she had already made an exploratory field trip to Venezuela to look for children that have both copies of the faulty Huntington’s gene — the homozygote. Now, at the October workshop, she and her team realized that the Venezuelan communities might hold the key to the markers.

“There have been few times in my life when I felt convinced that something was really right,” Nancy wrote later, “times when my heart has raced and leapt into my throat, times when I couldn’t sit still and wanted to race as fast as I could, laugh wildly or explode. I had this feeling at the end of the workshop.” She then planned a serious expedition to South America.

Blood Samples and Family Trees

In March 1981, Nancy and a team of seven American researchers and five Venezuelan researchers began their first expedition, traveling to San Luis, Barranquitas, Lagunetas, and other stilt villages along Lake Maracaibo. The goal was to continue looking for the homozygote, and to begin creating a family tree for the communities, taking blood samples to follow the inheritance of Huntington’s and enable the genetic linkage analysis.

To court the cooperation of residents, Nancy enlisted Dr. Américo Negrette, who had documented Huntington’s in the area two decades earlier and was trusted by the communities, along with Ramón Avila-Girón, whose film she’d seen at the conference, and other Venezuelan partners. Convincing the residents, however, would take more than some friendly, earnest faces. At an informal party in the barrio, Nancy shared in her pidgin Spanish that she too had Huntington’s — el mal, as the locals called it — and their goal was to find the cause and cure. The families were stunned to hear that this disease existed outside their towns. Some didn’t believe her. But Fidela Gomez, the team’s nurse, held up Nancy’s arm and walked her around the room, pointing out the biopsy scar on her arm. The gesture made them understand; they no longer doubted their visitors.

Alice Wexler joined on two of these trips and beautifully captured what their research looked like: kids darting around the meeting room while adults shooed them outside; walls covered with family tree charts peppered with Polaroid photographs and names; researchers rehearsing their questions in Spanish between check-ups. Children crowded the sidewalks, mimicking neurological exams, performing the “follow-my-finger” ritual just as they saw the neurologists do. Draw days were particularly intense — the team had 48 hours to obtain, package, and ship the samples to laboratories in Maracaibo, Caracas, or Boston.

Every year for the next decade, more or less the same team made the trip to the stilt villages. In the sweltering heat and humidity, they ran neurological tests, conducted questionnaires, and collected blood samples and personal information. Nancy played the role of “physician, nurse, ethnologist, psychologist, diplomat, photographer, neurologist, geneticist, and general all rolled into one.” By the end, the team had collected nearly 4,000 blood samples from healthy and sick individuals, and the family tree they created encompassed over 18,000 individuals. After some local investigation, they even managed to trace the origin of the mutated gene to a woman who was believed to have lived in Laguneta in the 1800s, aptly named Maria Concepcion.

Patterns in the Dark

The blood samples were shipped to James Gusella at the Massachusetts General Hospital in Boston and to Michael Conneally at Indiana University. After the October 1979 workshop, Gusella began testing RFLP markers on a small Huntington’s disease family lineage from Iowa. By the spring of 1983, they had tested eleven markers without success.

Then came the twelfth probe, called G-8, named after technician Ginger Weeks. When tested on the Iowa family, the results showed promise — the LOD score, a statistical measure of genetic linkage, was 1.8, about 63-to-1 odds that they were linked. This was good, but not high enough. The researchers were looking for a LOD score of at least 3 — 1000-to-1 odds — to be “definitively significant.”

Gusella decided to test G-8 on DNA extracted from the Venezuelan kindred, realizing that the sheer size of that family could provide the definitive proof needed. In late July 1983, as Gusella and his technician sat down to read the autoradiograms, they saw that the inheritance patterns “matched perfectly.” Individuals with Huntington’s disease in Venezuela consistently carried a specific DNA fragment pattern — a “C haplotype” — while those without the disease carried a different pattern. Conneally’s statistical analysis confirmed it: the LOD score was above 6. Better than a million-to-one odds.

Their paper published in Nature revealed that the Huntington’s gene was somewhere on the short arm of chromosome four. Gusella estimated that their G-8 marker lay at least four million base pairs away from the gene – close enough to create a predictive test that was 95% accurate for families with sufficient genealogical information. It was the first time this RFLP technique had successfully located a gene for an inherited disorder whose chromosomal location was completely unknown. James Watson later stated that this result helped launch the Human Genome Project.

The invention of the predictive test created a new ethical and emotional quandary: does one want to know whether they will get a debilitating and fatal disease for which there is no treatment or cure? The answer, it turned out, wasn’t simple. In the Wexler family, taking the test they had worked so hard to make a reality seemed like a terrifying idea. Their entire lives, Nancy and Alice felt that they would not get it, since their father kept confidently saying so. The “denial” the sisters had cultivated — now revealed as an intentional tool fostered by Milton — allowed them to live somewhat normally with the ambiguity. A test now would shatter that deniability. “The thought of learning that I carry the gene — that my brain is already deteriorating — is just too horrendous. I’m not sure I could go on.” Alice wrote in her diary in May 1984. Nancy was worried too: “If the test showed I have the gene,” she wrote in 1991, “would I continue to feel the happiness, the passion, the occasional ecstasy I feel now? Is the chance of release from Huntington’s worth the risk of losing joy?”

Milton urged them not to take the test, pointing out that the 5% chance of error added yet another layer of ambiguity. The test became the subject of many family arguments.

When tests started rolling out across the United States, Canada, and Great Britain, relatively few people took them. Alice Wexler decided to ask some test-takers why. One fifty-year-old man revealed, “My feeling was that I did have it, so it would not be that awful to find out for sure.” He was among several people she interviewed who always believed they would get it — a confirmation would erase the ambiguity and help them plan better, while a good result would change their lives. For them, it was upside either way.

Chromosome Walking with the Gene Hunters

While the G-8 marker discovery was an accomplishment, it was not the gene itself. That remained about four million base pairs away, and the task now was to physically traverse this distance along the chromosome, like searching every address in a neighborhood. To find the gene as soon as possible, at a January 1984 HDF workshop, Nancy persuaded researchers to form an unprecedented collaboration of six laboratories across the United States and Britain. Formally, they were known as the Huntington’s Disease Collaborative Research Group (HD-CRG). Informally, they were “The Gene Hunters.”

By early 1986, the group had narrowed the search to a region at the top of chromosome four, from the G-8 marker to the tip. The goal was now to find a marker on the other side, called a flanking marker, creating a “fence” around the target gene.

That turned out to be a lot trickier than expected. Contradictory data from Venezuelan siblings pointed at two different possible locations: the tip of chromosome four, or the inner region. Researchers focused their hunt on the tip region, suspecting it would be there, but it turned out to be a bizarre and intractable area. The tips of chromosomes are more inclined to recombination events during reproduction, scrambling up the statistical patterns through inheritance they were checking. After years of searching, in 1990, they realized that the gene wasn’t there. They’d been looking in the wrong place.

It had been seven years since the breakthrough marker discovery with still no gene to show. The frustration was starting to build up, but the camaraderie of the collective structure kept them going. As they began searching in the inner region, a clue surfaced. Housman had recently found the gene for myotonic dystrophy, where “a short sequence of three nucleotides that is usually repeated just a few times undergoes abnormal expansion,” mangling the protein’s functioning and causing the disease. At a January 1992 HDF workshop, he wondered if Huntington’s was caused by this same abnormal trinucleotide repeat.

The clue proved crucial. On February 24, 1993, James Gusella and his team identified the Huntington’s gene, initially named IT15 — “interesting transcript 15” — and later renamed huntingtin, or HTT. The “longest and most frustrating search in the annals of molecular biology” was finally over. In recognition of the unprecedented scientific cooperation, their paper, published in Cell a month later, listed its author as the Huntington’s Disease Collaborative Research Group, with the 58 members of all six labs credited underneath.

Located 3.5 million base pairs from the tip of chromosome four, the gene was enormous: 170,000-180,000 base pairs long. The normal gene contained a CAG triplet — CAGCAGCAG… — that repeated about 17 times on average, ranging from 10 to 35 normally5. The mutation that causes Huntington’s increases the number of repeats to more than forty, sometimes exceeding a hundred. The more repeats, the earlier the symptoms occur and the higher the severity.

Huntingtin, the protein encoded by the gene, is a massive 3000-plus-amino-acid protein, expressed in nearly all cells, particularly neurons. It acts as a traffic-controller, helping move vesicles and organelles and organizing other proteins to ensure they hum along on their cellular pathways. When the CAG repeat is abnormally long, the protein misfolds and gets abnormally cleaved, forming clumps that derail its function. This dysfunction devastates neurons, eventually causing the cognitive drift and macabre dance that are Huntington’s telltale signs.

For those who had dedicated their lives to the search, finding the gene brought overwhelming triumph, relief, and elation. Nancy was just about to leave for South America when she received the news. “It was incredible,” she shared. “I just started screaming at the top of my lungs: ‘We found the gene!’ I called my dad. I said, ‘Dad, we did it,’ and he started crying. I called my sister. It was just euphoric.” Milton, then 84, was thrilled to see this discovery in his lifetime. James Gusella and his team felt “pure joy,” thrilled that the search was over. David Housman did what he always does after a significant discovery: he went out and had a beer.

Finding the gene simplified the presymptomatic test — no family structure requirements, just a simple blood test. But, since there was still no cure, the psychological uncertainty around the test didn’t change all that much.

Nancy and Alice both decided not to take the test. Recent reports indicate that Alice, now 83, has not developed any symptoms. Nancy’s fate, however, unfolded differently. Today, at 80, Nancy Wexler lives with the disease that has haunted her family for generations. Her speech slurs. Her limbs jerk in random directions. When she walks, her gait has the same unsteady rhythm her mother’s once did. The movements are involuntary, relentless, unmistakable.

But her mind remains sharp. In 2024, she participated in an extensive interview with the Lasker Foundation, discussing her life’s work and the science she has championed for decades, talking about gene mapping breakthroughs to therapeutic possibilities. Through the involuntary twitching and trembling, her voice carries the same passion it has since the day they began hunting for the gene.

While the disease may choreograph its cruel dance through her body, it never controlled her life’s meaning. That summer day in 1968, when Milton told his daughters they faced a 50-50 chance, none of them could have known what would follow. Milton would create a new model for scientific collaboration, bringing together minds that would never have otherwise met; the Venezuelan families would trust outsiders with their blood and their stories; James Gusella and the Gene Hunters would persist through seven frustrating years after the initial breakthrough, finally finding the gene; and Alice would stand beside her sister through it all, documenting the human cost as Nancy orchestrated the science.

Nancy’s curse became a challenge, and then a coordinate on chromosome four. The huntingtin gene still lurks in her DNA, just as it did in her mother’s, her uncles’, her grandfather’s. But because of the legacy they all built together, it no longer lurks in darkness. Huntington’s disease is a problem with a location, a mechanism, and now an experimental treatment. Families around the world now have a choice her mother never did.

Thank you to and for feedback and edits.

In the early 19th century opera Der Freischütz (The Marksman), Max, a young hunter, in order to marry his beloved, must win a challenging shooting contest. Anxious and lacking confidence, he makes a deal with the Devil to receive magic bullets that always hit their targets.

In 1907, German physician Paul Ehrlich borrowed the term to introduce a new scientific idea. The “magic bullet” — or Zauberkugel — would describe the dream of medicine: to target and kill the sole cause of disease in a body without harming the body itself.

While several therapies now meet this standard, one rises above the rest in its range and impact: monoclonal antibodies. These lab-made antibodies — the little Y-shaped proteins used by your immune system to identify and fight off foreign substances — have saved or drastically improved tens of millions of lives fighting cancer, autoimmune diseases, infectious diseases, allergies, and more.

This is the story of how the monoclonal antibody went from the pipettes of experimental laboratories to the bioreactors of industrial production.

On a crisp winter morning at the Hôpital des Enfants-Malades in Paris, Émile Roux hung a bottle of pale horse serum over a child’s bed. It was taken from horses that had been injected with a weakened diphtheria toxin; the hope was that the “antitoxins” they produced could help the child — and the rest of the ward of sick children — fight diphtheria. By that same summer in 1894, Roux’s team had administered this treatment to 448 children. Their death rate was 24%, compared to the 60% at a nearby hospital that didn’t use the serum.

News of this astonishing success spread quickly. Soon, hospitals in other cities around the world began immunizing horses to harvest the same antitoxin. It dramatically reduced mortality, but a question hung over: what was this mysterious element coursing within horses’ serum that was helping fight toxins? Before anyone had seen the molecule, and years before these diphtheria experiments, Paul Ehrlich had already come up with a universal name for the element: Antikörper — or antibody.

Bleeding animals to harvest antibodies naturally had drawbacks. They could only get so many antibodies from a horse or rabbit or mouse. This meant lots of animals were needed. In large-scale productions, this could mean thousands of mice and hundreds of sheep, goats, or horses, often housed in dedicated farm-like facilities. Multiple immunization rounds across different animals created different antibody mixes, which meant no consistency between batches. And extracted cells could not be cultured long-term, so blood had to be drawn repeatedly.

Even with all this effort, the resulting molecules weren’t ideal for therapeutic use. Antibodies are produced by B-cells, one of the two crucial fighting arms of your immune system. Different B-cells produce different antibodies. Some of the harvested antibodies target the disease, but others can target healthy proteins in the body, introducing the chance for unpredictable side-effects. These antibodies harvested from animals were “polyclonal,” making them more of a shotgun blast than the magic bullet that researchers were looking for.

To make a single, defined antibody on command, you’d need to take one B-cell that makes it and grow the B-cell indefinitely in a dish, so the B-cell keeps dividing and secreting the same antibody. Unfortunately, normal B-cells don’t keep dividing in vitro, so even if you found the right cell, it would die out rather than become a factory line.

Then, in 1975, two scientists found an ingenious solution to this B-cell culturing problem. Georges Köhler, a German biologist, and César Milstein, an Argentinian biochemist, discovered that if you fused a B-cell with a cancerous myeloma cell, you could get the antibody-producing power of the B-cell with the immortality of the myeloma cell. Now, rather than having to repeatedly draw blood from animals, they created tiny cellular machines — “hybridomas”, as they called it — capable of making identical antibodies, indefinitely.

In one fell sweep, their invention leapfrogged a number of the key drawbacks of producing antibodies through animal immunization: specificity (antibodies were now monoclonal, produced from a single B-cell clone), supply limits (now technically unlimited), and labor intensity (drawing blood was replaced with pipetting cells).

While promising, scaling and improving yields of hybridomas still remained a challenge. In those early days, hybridomas were cultured inside the bellies of mice1. It produced small yields, giving mice cancer raised ethical issues, it took four to six months, and was still quite labor-intensive. It was also incredibly inefficient: 99% of the hybridoma cells died during fusion, or never fused at all.

Worse, however, was that when these mouse-derived antibodies were given to humans in early clinical trials, patients’ often had strong immune responses. The drugs were instantly neutralized or caused allergic reactions, ranging from mild rashes to life-threatening events. Researchers gave it a straight-arrow name: the Human Anti-Mouse Antibody reaction, simply known as HAMA.

Fixing these issues required another paradigm shift; or rather, borrowing from one. Around the same time Köhler and Milstein frankensteined the “hybridoma”, another revolution was at the cusp of taking the world of biology by storm: recombinant DNA. Over some late night sandwiches at a conference in Hawaii, Herb Boyer, a biochemist at UCSF, and Stan Cohen, a geneticist at Stanford, combined their research to hypothesize a molecular “cut-and-paste” function, potentially giving researchers the power to take a desired gene from one species and insert it into the genome of another. The company they founded on the basis of this technique — Genentech — would go on to become an expert at editing the genes of microorganisms to produce human proteins. Working with the pharmaceutical giant Eli Lilly, they used recombinant DNA to induce bacterial cells to produce human insulin, which they marketed as Humulin. Not only was bacteria-produced insulin safe, it was superior to the animal-derived insulin typically used by diabetics. They also managed to figure out how to quickly produce it in enormous quantities.

Could bacterial cells be used to produce monoclonal antibodies in enormous quantities too? Sadly, no. Bacterial cells simply did not have the molecular machinery to produce these larger, more complex proteins2. For that, mammalian cells would be required. But which one?

In 1948, during the height of the Chinese Civil War, an American researcher raced 20 Chinese hamsters across war-torn China, driving 11 hours past Communist patrols to get them on the last Pan Am flight out of Shanghai. In 1957, Theodore Puck, an American geneticist, would use cells from one of these hamsters to derive a cell line that would revolutionize medical research and production of biologics. Prior to their consideration for monoclonal antibodies, Chinese Hamster Ovary (CHO) cells had already been used successfully in genetics and cancer research.

CHO cells had unique biological and practical advantages: they were easier to modify for human therapeutic use, and did not cause immune responses. They were also adaptable to better manufacturing techniques. Rather than being limited to growing on a flat surface, the cells could be cultivated by the trillion in suspension cultures, floating throughout enormous vats of nutrient-rich liquid like tiny biological submarines — a method perfectly suited for industrial scale-up.

CHO cells could also be further adapted for this suspension culture environment. Rather than depending on fetal bovine serum as the nutrient medium (which could result in batch variation and contamination risks), they could be adapted to grow in a serum-free environment, using a precisely formulated mix of nutrients instead3. They are also remarkably willing to accept new genetic instructions, which not only made it simple to add the gene to produce the desired antibody, but also enabled the use of a gene amplification system to multiply that gene hundreds of times within the cell's DNA, dramatically boosting its production capacity.

All these attributes made it pretty much the ideal cell line to produce human antibodies for human therapeutic use. Now, hybridomas could provide the genetic blueprint for specific antibodies, and CHO cells, suspended in steel tanks, could be the molecular factories. Add to this a downstream platform tasked with purifying the final product from the raw cellular broth (a refinery, if you will), and we have a complete system to manufacture monoclonal antibodies. All that was left was to scale it up.

Scaling would turn out to be very challenging. Unfortunately, in biology, simply making things bigger doesn't mean more output. To go from laboratory flasks to today’s 20,000 L bioreactors, and from antibody yields of ~0.1 g/L in the 1980s to pushing 16-17 g/L today, required a tremendous amount of ingenuity, effort, and engineering prowess.

Every problem that came up required a unique solution. Take bioreactor size, for instance. As tanks get larger, volume increases exponentially, but there’s proportionally much less surface area for oxygen to enter. To prevent cells in the middle and bottom of the bioreactor from starving of oxygen, engineers installed mixing systems to diffuse oxygen more quickly, as well as air spargers to introduce oxygen through tiny openings in the bottom or sides.

These additions transported oxygen to the places that needed it, but it introduced a new problem: the mechanical forces from all that mixing started damaging — stretching, squashing, even tearing apart — the CHO cells. So, engineers fixed this by switching to broad paddle-shaped impellers that moved liquid more slowly, and installing fine-pore spargers that released hundreds of tiny air bubbles rather than the more potentially damaging large ones. These innovations were just the first step in a long journey to push yields beyond the initial ~0.1 g/L baseline of the 1980s.

The 1990s saw the next breakthrough. Rather than dumping all the nutrients into the bioreactor at once, scientists developed feeding strategies that continuously supplied fresh nutrients while the cells were growing. Their innovation, known as fed-batch processing, extended cells’ productivity by preventing feast-or-famine cycles and avoiding toxic accumulation of cellular waste. Engineers also began introducing sensors that precisely measured critical parameters like oxygen and pH levels. With these, titers — i.e. the concentration of antibodies in culture — rose to ~1-2 g/L — a 10-20-fold improvement — and allowed researchers to scale up to ~10-kL bioreactors.

The 2000s brought some optimizations to the processes introduced in the 1990s. The chemical recipes that nourished cells got much better, eliminating most variability from feeding practices. More sensors were added, providing more real-time feedback. And virus filtration systems in the downstream process — i.e. the refinery segment — improved product safety without sacrificing yield. With these, titers rose to ~3-6 g/L — a 3-6-fold improvement — and bioreactors scaled up to 25-kL fed-batch systems.

The 2010s saw another paradigm shift. Waste accumulation inside the tanks meant that fed-batch cycles generally lasted 14 days, after which tanks needed to be cleaned and sterilized. A new technique revolutionized this by constantly replacing the spent medium with a fresh one using cell-retention devices — microscopic sieves that kept the cells inside while draining the waste. Now, rather than 14-day cycles, cell cultures could run these perfusion cultures for weeks or even months without stopping.

Even with these longer cycles, tanks still eventually had to be cleaned and sterilized, which would mean up to a week of downtime. Another optimization wanted to skip cleaning altogether. Single-use bioreactors — essentially pre-sterilized plastic bags — meant that this process could now take hours; simply throw out the plastic bag and put in a new one. With all these developments, titers rose to ~6-12 g/L, another 2-4-fold improvement.

Today’s systems have squeezed out even more efficiency gains. For one, they can now run nonstop. Continuous upstream perfusion can be combined with continuous downstream purification, so cells can be constantly pumping out antibodies. Add to this some more advanced sensors with automated feedback control, machine learning algorithms, and other incremental improvements, antibody titers in optimized and intensified process can rise up to ~16-17 g/L.

To translate that into steel and liters: at a 20,000 L production scale, a 10-15 g/L run produces roughly 200-300 kg of antibody before purification. Across the industry, annual output has climbed from about ~10 metric tons in 2013 to roughly 25 metric tons in 2019 and ~30 tons in 2020. During COVID, the sector produced ~30+ tons of antibody products for COVID alone. Today’s global capacity is estimated at ~80–100 tons per year, enough for tens of millions of doses depending on strength.

The 100-200-fold productivity gains to enable this kind of production scale were a result of solving many complex engineering puzzles as they came along — four-plus decades of paradigm shifts and innovations stacked upon innovations. Köhler and Milstein's hybridoma — the ability to continually produce a specific antibody — was the biological spark; recombinant DNA prevented that spark from being extinguished by the human immune system; and CHO cells in bioreactors provided the factory. But it took four decades of relentless engineering — solving puzzles of oxygen flow, nutrient delivery, waste removal, and more — to scale that factory from a workbench to a city block.

True medical revolution generally requires more than a single moment of genius. In the 1970s, extracting these monoclonal antibodies was solely the subject of laboratory experiments, largely with mice. Today, it represents the fastest-growing class of therapies worldwide, with a global market size of approximately $250bn as of 2024, and is expected to grow to $700-800bn by 2033. The story of monoclonal antibodies required bridging the distinct worlds of immunology, genetic engineering, and process engineering, creating a seamless production line where every step, from the cell's first division to the final purification, has been meticulously optimized.

As far as I’m aware, there was no deal with the Devil for these magic bullets. Just thousands and thousands of scientists, engineers, researchers, doctors, and many, many more around the world working very hard to solve complex scientific, engineering, and social problems. Recent clinical trials show that monoclonal antibodies have expanded to serve new disease areas, and every year, millions more will benefit from this revolutionary medicine. It is thanks to these efforts that millions of patients — cancer patients, COVID-19 patients, individuals with autoimmune diseases, and more — get to live a little longer, or a little more comfortably.

Thanks to , and for feedback and edits.

This is my first essay as part of the Roots of Progress Blog-Building Intensive! Check out the others in the program.

First attempts in a new field of medicine rarely go according to plan. On September 14, 1990, Dr. William French Anderson and his team at the National Institute of Health (NIH) performed the first official gene therapy trial. The patient, a 4-year-old Ashanti deSilva, suffered from a rare genetic disease called adenosine deaminase (ADA) deficiency, a form of severe combined immune deficiency (SCID). Children with ADA-SCID rarely make it to adulthood; the lack of a functional immune system makes any illness potentially lethal. To make up for the deficiency of this crucial enzyme, Ashanti had been receiving ADA injections since she was two, but the effectiveness of this treatment usually declines fairly quickly, and by age four, she was no longer responding to it.

Ashanti’s parents, Raj and Van DeSilva, felt like they had run out of options for their daughter. They put all their hopes on this trial. Anderson and his team extracted white blood cells from Ashanti’s body, inserted a functional copy of the ADA gene into these cells using a retroviral vector, and then infused the genetically modified cells back into her body. Remarkably, this worked. Ashanti’s immune system function improved over the next few months; her T-cell count rose dramatically, and she no longer constantly fell sick. It was not a one-time cure — she still needed regular infusions every two months to maintain her health — but to her parents her recovery was nothing short of miraculous. Ashanti could begin living a normal life for the first time; going to school, for one, was no longer a life-threatening affair.

Thirty-five years since that first success, a very similar story swept the world. On May 15, 2025, Kyle Junior Muldoon — better known to the world as baby KJ — made headlines after the announcement that he had been successfully treated with the first personalized, CRISPR gene-editing therapy. Just nine months earlier, in August 2024, mere days after birth, KJ was diagnosed with carbamoyl phosphate synthetase 1 (CPS1) deficiency, an extremely rare and often fatal genetic disorder that prevents the body from breaking down ammonia. What followed was an extraordinary race to create a cure just for him: within days, his DNA was sequenced to find the specific mutations (two, in his case) in the CPS1 gene. Then, a cutting-edge base-editing therapy was designed tailored to these mutations. By months four and five, preclinical safety and efficacy testing was underway in mice and monkeys, and by month six, the FDA approved this single-patient product within a week of the application.

After receiving two infusions of his therapy — one in February 2025, and another a month or so later — he quickly began getting healthier. He could tolerate more protein, started gaining weight, needed fewer medications, and could achieve simple, regular baby activities, like sitting upright. On June 3rd, 2025, nine months after he was born, KJ was discharged from the Children’s Hospital of Philadelphia. His parents took him home for the first time.

These two extraordinary stories, three and a half decades apart, share a lot in common: parents grasping onto the last thread of hope for their child, the magical idea of fixing previously untreatable afflictions at the very core, teams working urgently for months to develop a cure, and, ultimately, a chance for a child to live a regular life.

But there are also some key differences that reflect the progress the field of cell and gene therapy has gone through in that time: new technologies, updated regulations, better development speed, safety, and efficacy, and a broader number of target diseases. Ashanti’s trial took two years of rigorous reviews from application to approval; KJ’s took two weeks. Ashanti’s therapy used a retroviral vector to deliver the functional gene, a delivery mechanism that would later be culprit in accidentally inserting genes away from its target (a possibly devastating outcome known as insertional mutagenesis). KJ’s, on the other hand, used lipid nanoparticles — these incredibly minuscule fat-bubble suitcases — to transport mRNA, which is then translated into a base-editor protein that edits — as if correcting a spelling in a word processor — only the specific mutation. Ashanti’s therapy was ex vivo: cells were extracted and modified outside the body; KJ’s therapy was in vivo: his body performed the modifications itself. Ashanti’s therapy was not personalized; the same product was used for all ADA-SCID patients. KJ’s was personalized to his exact mutation.

KJ’s story was celebrated across the world. This incredibly rapid, precise, effective therapeutic future should, after all, imbue a grand sense of excitement. His triumph over impending death is an astonishing demonstration of what 21st century medicine can achieve when cutting-edge science, experienced and compassionate medical staff, and proactive regulators come together, driven with a sense of urgency. These “elite institutions with complementary superpowers” — as Fyodor Urnov, Scientific Director at the Innovative Genomics Institute at UC Berkeley and one of the key characters in the baby KJ success story puts it — gave a boy with almost no hope, and his family, the chance at a regular life.

And yet, this triumph reveals a troubling paradox. We now find ourselves at an inflection point where the science to perform such miracles, to cure rare and even ultra-rare genetic diseases — like spinal muscular atrophy (SMA), hemophilia A, even restoring vision — is not only advancing rapidly but is demonstrably here. Regulatory pathways, like with the rapid approval for baby KJ, have even evolved to support these breakthroughs. But, despite saving many lives, many therapies born from these advances don’t ever get commercialized, and if they do, many are unable to translate into sustainable businesses.

So instead, we are seeing bankruptcies, downsizing, plummeting stock prices, and on-the-cheap acquisitions. Bluebird Bio, once a leader with three FDA-approved gene therapies, sold itself to private equity for around $30 million earlier this year, a 99.7% decline from their all-time-high $10b valuation. In February this year, Pfizer abandoned Beqvez, its FDA-approved hemophilia B gene therapy, citing “limited patient and physician interest”, and effectively exiting the gene therapy field altogether. Editas and Beam, two highly-acclaimed public companies in the gene-editing space, have seen their share prices plummet by 90% since their 2021 all-time highs. These troubling commercial and financial setbacks are scaring away the crucial investment needed for developing future cures or commercializing existing ones.

This is the tragedy-to-be: science and technology offer a cure, but we risk a future where these life-saving treatments remain undeveloped or inaccessible, simply because we haven't yet built a system that can support them. In economic terms: cell and gene therapies suffer from a market failure.

The immediate question is: why? Why aren’t cell and gene therapies — evidently capable of miraculous cures — turning into successful businesses? What are the factors that prevent these products from escaping the “valley of death” — flourishing beyond the initial scientific breakthrough or successful clinical trials? Answering that might give us hints at asking the larger questions. How might we fix it? What might a system that incentivizes these rare-disease therapies look like? What are the regulations and funding models — designed for a world of off-the-shelf, repeated use medicine — that can support developing personalized therapies? And most importantly, how can we deploy these life-saving technologies to the people that need it?

Perhaps these questions surrounding the science, the treatment, the regulations, and the economics are best illustrated with an example. After going through multiple FDA programs — Fast Track, Orphan Drug, RMAT, and a Rare Pediatric Disease priority review voucher — in January 2024, Casgevy became the first FDA-approved gene therapy that uses CRISPR-Cas9. It is a treatment for sickle-cell disease and beta thalassemia, diseases caused by a genetic defect that produces hemoglobin — the shipping-container-like protein in red blood cells that ferry oxygen across the body — with an irregular shape (sickle-shaped, giving SCD its name) or in insufficient amounts (beta thalassemia).

The treatment is ingenious, and makes use of one of the wonders of the human body. Before birth, fetuses in the womb use fetal hemoglobin to ferry oxygen across the body. After birth, the fetal hemoglobin gene is turned off, and replaced by adult hemoglobin. Since patients with SCD have deformed adult hemoglobin, one possible option was reactivating fetal hemoglobin instead. Casgevy works by extracting blood stem cells from bone marrow, knocking off the regulator that has switched off fetal hemoglobin production, and then infusing the stem cells back in the patient. The restarted production of fetal hemoglobin compensates for the faultiness of the adult ones.

This isn’t a complete cure, but it significantly reduces painful episodes (the sickle-shaped red blood cells can get stuck in capillaries, causing intense pain known as a vaso-occlusive crisis) as well as the need for blood transfusions. And, since it’s a single treatment that uses the patient’s own cells, there’s no risk of graft-versus-host disease — that terrible affliction where donor immune cells attack the recipient — and no repeated need for immunosuppressants.

For nearly every patient that has received it, Casgevy has been an enormous success. In the final clinical trials before approval, over 100 patients had been treated. For sickle-cell disease, among patients with at least 16 months of follow up, 92% (36 of 39) were free of vaso-occlusive crises for at least 12 consecutive months. For beta thalassemia, 94% (49 of 52) didn’t require a blood transfusion for at least 12 consecutive months. Early data tracking post-approval treatments had similar rates for the same outcomes.

Jimi Olaghere, a young Atlanta resident, was one of the patients enrolled in these trials. He shared his experience with Casgevy in an op-ed for MIT Tech Review: “I started to experience things I had only dreamed of: boundless energy and the ability to recover by merely sleeping. [...] I gained the confidence that sickle-cell disease won’t take me away from my family, and a sense of control over my own destiny.”

For patients like Jimi, this potentially curative, one-time infusion of their own, edited cells allowed them the chance to live a more regular life. But there’s a catch, and a little more to this tale.

For one, Casgevy has an incredibly high price. The one-time treatment is set at $2.2 million, and this doesn’t include additional expenses such as pre-treatment evaluation, the hospital stay for stem-cell collection, and any follow-up care.

If one does decide to go ahead with the treatment, the entire process is gruelling, and can take up to 5-6 months. First, the patient is evaluated — screened for infections and certain medications — at an authorized treatment center. Then, the patient receives medications to mobilize hematopoietic (blood) stem cells from the bone marrow into the bloodstream. Over the next few days, blood is repeatedly drawn out and the specific stem cells needed — CD34+ cells — are separated and collected, which is then shipped in a sterile bag to a specialized manufacturing facility. There, the cells are further purified and exposed to the CRISPR-Cas9 gene-editing complex through electroporation, a process where an electric pulse opens up the pores in the stem cell membrane, allowing the molecular complex to enter. Once inside, the CRISPR-Cas9 targets and edits a specific region of the BCL11A gene, which reduces the gene’s expression, and in turn, increases the production of fetal hemoglobin (once inside the body).

These modified stem cells are then mixed with a cryopreservative solution and frozen in liquid nitrogen for storage. A small sample undergoes extensive testing for viability, purity, potency, sterility, and safety, which can take months since the cells need to be cultured. Once all the tests have passed, the final product is transported in a special cryoshipper back to the treatment center.

Before receiving their edited cells, the patient has to undergo chemotherapy to clear out their existing bone marrow cells. Once depleted, the modified stem cells are thawed and infused into the patient’s bloodstream, and they’re kept in the hospital until their new cells engraft and their blood count recovers.

This entire Casgevy process illuminates why such a therapy can be so expensive. It requires a phenomenal amount of resources: specially trained medical staff, temperature-controlled transportation, sophisticated gene-editing capabilities, and comprehensive safety checks. There are no economies of scale; each patient’s therapy is a separate campaign. And there are clear bottle-necks and failure points. Cells begin deteriorating within hours of collection and degrade rapidly if temperature deviates during transport (it needs to be kept at 2-8°C for unedited cells; -150°C for edited cells). And a significant chunk of time is spent waiting for the stem cells to culture during testing. A single misstep or accident could undo months of progress.

This operational complexity and fragility makes providing Casgevy a difficult business to operate and scale. Despite their clinical successes and blockbuster announcement, sales figures in the last quarter of 2024 and first quarter of 2025 have fallen short of analyst expectations1.

The need for specialized treatment centers and specially-trained staff were two key reasons cited for their lagging pace. Another one was simply the slow uptake of patients. Even after regulatory approval, it took many months before the first patients were treated; ten months after approval, only eight patients had completed treatment (although 90 had started the collection process). The high price poses a major barrier to patients — insurers are cautious about paying, often requiring complex negotiations, outcomes-based agreements, or risk-sharing models before agreeing. And, despite the clinical trial successes, it can be easy to forget that this is still a difficult, medically intensive journey for the patient, requiring chemotherapy, hospitalization, and months of recovery. Along with complex manufacturing, these patient-specific factors — eligibility, willingness, location — and a complex reimbursement landscape make it hard to predict revenue. Thus, despite being a breakthrough product, the path to profitability for Casgevy seems uncertain.

These “bench to bedside” hurdles are not unique to Casgevy. Many FDA-approved cell and gene therapies have similar operations and scaling hurdles. All seven cancer-fighting CAR-T therapies, which include Yescarta for B-cell lymphoma and Kymriah for acute lymphoblastic leukemia (ALL), require similar steps — collect T-cells, modify with vectors, multiply, and reinfuse — and thus potentially have the same scaling drawbacks. As do the two other therapy options for sickle-cell disease and beta thalassemia — Lyfgenia and Zynteglo, respectively — just swapping CRISPR with a lentiviral vector. Typically, it’s the allogenic options — cells taken from donors — that skip the collection and gene editing or delivery steps, but it has other drawbacks of its own, like donor-patient matching, and, of course, are not personalized.

Operational complexities, multi-million-dollar price tags, challenging reimbursement landscapes, and a slow and uncertain path to commercial viability represent systemic issues that threaten the broader promise of cell and gene therapies. But the problems identified also hint at possible solutions. From the Casgevy example, we can garner three primary headwinds: manufacturing hurdles, cooling investment, and reimbursement struggles. In Part Two, we’ll take a closer look at these three facets for the industry as a whole. Then, we’ll explore some potential solutions that either fix the issue directly, or jump past it altogether, particularly those brought up in the June 5, 2025, FDA Roundtable on Cell and Gene Therapy.

One day in the early 1980s, a teenager from Higashiosaka, a bustling industrial city east of Osaka proper, accompanied his father to the hospital. Shozaburo, a middle-aged engineer, ran a family-owned factory with the help of his wife, designing and manufacturing components for sawing machines, the same factory his father ran before him. That day, he had a small accident – nothing critical, but it did require a blood transfusion.

Everything seemed fine for a while, but then one day he suddenly fell ill. The symptoms suggested hepatitis, but tests for hepatitis-A and hepatitis-B came back negative. This mystery illness slowly caused his condition to worsen; soon, a liver cirrhosis crept in. Suspecting his final days were approaching, he told his son that he should not take over the family business like he did, and he should become a doctor instead.

The son, then in his early twenties, looked up to his father immensely, and decided to follow his wishes and become a doctor. Astonishingly, his father lived to see him receive his M.D. from Kobe University, and even received medical treatments — injections and IV drips – from him while he was a resident at the Osaka National Hospital. But, sadly, a year later, in 1988, his father passed away, a fateful moment that would change the way a 26-year-old Shinya Yamanaka saw medicine, and his role in it.

Shinya would choose to specialize in orthopedic surgery, but would soon realize that he was not cut out for it. His hands lacked the dexterity to properly conduct surgery, and he routinely struggled to complete operations in a timely manner. His supervisors discouraged him from continuing on this path.

Feeling powerless after being unable to help his father despite his medical degree, Shinya too felt like this path wasn’t for him. A year after his father’s death, the virus that caused it was identified in the United States: hepatitis C. He realized, then, that you cannot treat what you don’t know, and found himself getting pulled in a new direction: medical research.

Trading the scalpel for the pipette, Shinya’s path led to a pivotal postdoctoral fellowship at the Gladstone Institute in San Francisco. There, he mastered the techniques of gene editing and embryonic stem cell culture, sparking a curiosity that would define his future research: how do these remarkable cells maintain their undifferentiated state?

After returning to Japan, Shinya Yamanaka struggled to find the funding and resources to continue the research he started. He nearly gave up and returned to medicine, but the breakthrough discovery of isolating human embryonic stem cells reignited his passion, illuminating the potential for curing patients using their own cells. When he got the opportunity to start at the Nara Institute, he decided to work on understanding these stem cells. While most labs explored how stem cells become specialized, Yamanaka posed a far more audacious question: can a differentiated cell return to its stem-cell state? Can you go the other way — can you turn back biological time?

Subscribe now

The pluripotency puzzle

At the heart of this emerging field lay a cell of almost mythical potential: the embryonic stem cell. Sourced from the inner cell mass of a nascent embryo just days after fertilization, they are nature's primordial building blocks, the ultimate biological blank slate. They possess an extraordinary, almost magical, property called pluripotency — the potential to differentiate, to transform, into virtually any specialized cell type the body requires, from rhythmic heart muscle cells and complex neurons to insulin-producing beta cells or protective skin. Understanding, and perhaps replicating, this remarkable adaptability became a major goal of modern biology.

Yamanaka's audacious “reversing biological time” question didn't arise in a vacuum. The scientific world had already been buzzing with possibilities and grappling with limitations surrounding ES cells. The birth of Dolly the sheep in 1996, cloned from an adult udder cell, was a landmark moment. It dramatically proved that the genetic instruction manual in a specialized adult cell wasn't permanently fixed; it could theoretically be reset to generate a whole new organism. This fueled excitement about the potential of human ES cells. However, ES cell research, particularly in the United States, became deeply mired in ethical controversy because deriving these cells typically involved the destruction of human embryos. This led to significant restrictions on federal funding and spurred many scientists, including Yamanaka, to seek alternative ways to harness the power of pluripotency without relying on embryos.

Beyond Dolly, other scientific clues hinted that reprogramming was possible. Researchers had demonstrated that transferring the nucleus (the cell's command centre containing DNA) from a somatic cell into an unfertilized egg cell could reset its developmental clock. Similarly, fusing an adult cell directly with an ES cell could sometimes coax the adult cell's nucleus into behaving more like an embryonic one. These experiments showed that something within eggs and ES cells possessed powerful reprogramming capabilities. But what, exactly? It remained largely a black box of unknown molecules. Yamanaka's hypothesis was a curious leap: perhaps this reprogramming wasn't just a vague cellular influence, but could be achieved by introducing a specific, defined set of genes — the very genes that normally maintain the ES cell state.

The experiment

Shinya Yamanaka assembled a team at the Nara Institute to test this theory. The premise of the experiment was simple: for an ES cell to maintain its undifferentiated state, it requires coordinated expression and restriction of certain genes. Pick the right combination of transcription factors — the proteins that latch onto genes to either express or restrict them — and the embryonic stem cell will stay as is, and not differentiate into a specialized cell, like a skin cell to protect your body or a red blood cell to shuttle oxygen around. After differentiation, these “ES maintenance” genes are usually turned off and “locked away”, and other genes that the specialized cell needs to function are then expressed. The question was: if those maintenance genes are “unlocked” and turned on again, will a specialized cell reprogram itself to its ES cell state?

To test this hypothesis, Shinya Yamanaka and his industrious post-doctoral researcher Kazutoshi Takahashi designed a series of clever experiments. They crafted a list of 24 candidate genes that were known or thought to be important in maintaining ES cell state. Their plan was to insert these genes into mouse embryonic fibroblasts (MEFs) — the precursor cells in fetal mice that differentiate into skin, muscle, and bone cells — and check if these induced the stem cell state.

To check whether the transformation was successful, they needed some kind of cellular signal. This required clever engineering: they found a gene that is only active in embryonic stem cells but not necessary for its functioning, named Fbx15, and modified it by combining it with another gene, β-geo, that makes the cell resistant to an antibiotic called G418. This way, if the cell is in an embryonic cell state, it will express the Fbx15 gene, which, now attached with β-geo, will create proteins to neutralize the G418 antibiotic. So now, embryonic stem cells from mice grown with this modified Fbx15 gene could survive very high levels of the antibiotic, but regular, differentiated cells would die even at normal levels. Thus, even a partial activation of Fbx15, indicating a partial move towards ES cell state, would result in the cells being resistant to the antibiotic (at normal levels), and thus distinguishing between ES cells and regular cells. The cells that survive will have gone back in time.

With the signalling mechanism sorted out, they began testing out which of the candidate genes might induce this state. First, they introduced each of the 24 candidate genes into mouse embryonic fibroblasts one by one, through a technique known as retroviral transduction, a method of delivering genes into cells using modified retroviruses1. But when they grew the cultures for each of these, none of the colonies were drug-resistant, suggesting that a single gene may not induce ES cell state.

If they didn’t work individually, maybe adding a combination of genes might? Yamanaka and Takahashi jumped to the other extreme: they added all 24 genes together. This time, after 16 days, small specks started to emerge around the dish: colonies. When inspected under a microscope, some of these cells even transformed their morphology, going from the spindle-like structure of MEFs to the round globules of ES cells, with “large nucleoli and scant cytoplasm.” They even started proliferating at the same rate as ES cells. The skin cells seemed to have literally shape-shifted. They called these cells induced pluripotent stem cells, or iPS cells for short, and labelled these specific ones iPS-MEF24.

This stunning result indicated that they were on the right track. Converting a differentiated cell back into a stem cell might indeed be possible. Now, they just needed to find out which of the 24 genes were actually inducing the ES cell state. Working backwards, they began taking out genes one by one; they hypothesized that the ones that stopped colony growth would be critical. Using this elimination method, they were left with 10 genes, and once they combined just these 10 genes — labeled iPS-MEF10 — more cell colonies grew than when they used all 24.

They then repeated the gene pruning process again. Removing genes one by one from the 10, they found that, incredibly, there were just four genes without which colonies wouldn’t grow properly: Oct3/4, Klf4, Sox2, and c-Myc.

Pruning the factors further didn’t work; they found that no combination of two factors would induce colonies, and using three factors sometimes yielded a few , but these were either unstable or looked abnormal — flatter, or rougher and spikier, than true ES cells. Thus, just three factors might initiate the reprogramming journey, but the specific combination of all four was essential for complete, stable transformation into an ES-like state.

So iPS cells using four factors mimicked ES cell morphology and activated the engineered Fbx15 marker gene. But was this resemblance more than skin deep? The critical question remained: were these induced cells truly pluripotent, possessing the defining capabilities of embryonic stem cells?

Answering this required moving beyond appearances and diving into rigorous molecular and functional validation. Real ES cells have characteristic gene expression patterns and the unique ability to differentiate into all cell types; Yamanaka and Takahashi now needed to subject their induced cells to a battery of tests to see if they truly measured up.

RT-PCR

The first test Yamanaka and Takahashi did was an RT-PCR to examine whether the key ES cell “marker” genes were actually being expressed2. The test revealed that in some clones, a majority of the markers that identify ES cells were switched “on”. This can be seen by the indicator bands, with the intensity corresponding to gene expression levels.

Chromatin immunoprecipitation

Next, they performed a test called chromatin immunoprecipitation (ChIP), which is a technique that’s used to study how proteins interact with DNA in living cells. The test helps researchers identify specific DNA regions bound by proteins like transcription factors or modified histones, providing insights into gene regulation and epigenetics3.

To understand what this test reveals requires a review of how DNA is arranged in a cell. DNA isn’t all laid about in a straight line, but wraps around proteins called histones like threads on a spool, creating a complex known as chromatin. When DNA is wrapped around tightly, the genes in that portion are silenced, and when it's packed loosely, genes can be active. Acetylation — adding an acetyl group to histones — loosens its grip, enabling gene activation, and dimethylation — adding two methyl groups to histones — tightens its grip, silencing genes. Thus, cells can use acetylation and dimethylation to control which genes are "on" or "off" without changing DNA.

The analysis revealed a crucial change in the iPS cells at the control regions (promoters) of key pluripotency genes like Oct3/4 and Nanog: histone acetylation had increased, while dimethylation decreased. This meant that the chromatin structure was physically relaxed, making these essential genes more "open" and primed for activation. The epigenetic “locks” were literally loosened.

DNA microarray analysis

A DNA microarray is a tool that also measures gene expression levels, but does so for thousands of genes simultaneously. This allowed the researchers to compare gene expression levels for thousands of genes for ES cells, iPS cells, and MEFs, and assess their similarity. Both a statistical analysis (Pearson correlation4) and the visual patterns — the red-green checkerboard — reveal that iPS cells are clustered closely to ES cells, and both distinct from MEFs. Yet the patterns weren't identical, confirming iPS cells were similar, but not exactly ES cells.

Histological analysis

The next important aspect that Yamanaka and Takahashi wanted to check was whether the iPS cells they created actually do what stem cells do: differentiate into a diverse set of specialized cells. They tested this by injecting the iPS cells into immunodeficient “nude” mice5 — and checking whether they develop into a teratoma, a tumor that includes a disorganized collection of tissues from all three germ layers: the ectoderm, mesoderm, and endoderm. After injecting their cells under the mices’ skin, a histological examination showed that they had indeed differentiated into cell lines starting from all the three germ layers; they found neural tissue (which comes from the ectoderm), cartilage (from the mesoderm), and columnar epithelium (cells that line the digestive tract, which come from the endoderm). Their induced pluripotent stem cells were capable of differentiation.

The researchers also decided to do some in vitro tests. All iPS cell lines were cultured in non-coated plastic dishes, where they cannot attach to the bottom, which forces them to form a spherical cluster of cells — in this case, embryoid bodies, which mimic the early stages of embryonic development. When grown in tissue culture dishes, where they can stick to the bottom, the iPS-MEF10 and iPS-MEF4 lines initiated differentiation, but the iPS-MEF3 cells did not. All cell lines confirmed the in vivo experiments.

Using adult skin cells

Seemingly satisfied with the results so far, the pair of researchers now went to change something that’s been constant from the start. Rather than test on mouse embryonic fibroblasts — cells derived from early embryos that retain some developmental flexibility — they decided to ask whether you could induce pluripotency from a fully specialized adult cell. In this case, they used tail-tip fibroblasts, the skin and tissue cells from a mouse’s tail. They injected the same four factors into four 7-week-old male mice, as well as one 12-week-old female mouse which continuously expressed green fluorescent protein. From the male mice, they established a line of iPS cells, which they called iPS-TTF4, and from the female mouse, they established 6 iPS cell lines (called iPS-TTFgfp4).

Under a microscope, these cell lines were indistinguishable from ES cells. Similar validation tests were performed on these adult-derived iPS cells: RT-PCR confirmed the expression of key ES cell markers, while teratoma assays showed their ability to differentiate into all three germ layers, mirroring the results seen with embryonic-derived iPS cells. They also injected the gpf version into mouse blastocysts, some of which developed in glowing mice embryos.

More tests

Incredibly, both mouse embryonic fibroblasts and adult mouse skin cells are undeniably capable of turning into pluripotent stem cells; it’s possible to reverse biological time. This itself would be an incredible discovery, but the researchers didn’t stop there. They wanted to find out in more detail what the differences were and whether they were crucial or not. They wanted to paint a portrait of this new cell.

Yamanaka and Takahashi performed more tests:

• A real-time PCR test revealed that the introduced factor genes were highly active, yet total protein levels mirrored those in ES cells, suggesting sophisticated internal controls managing the over-expression.

• A Southern blot analysis showed that each iPS clone had a unique pattern of where the four factors were inserted into the genome. The different insertion locations confirmed that the iPS cells were independently generated and not duplicates of one another.

• Karyotyping analysis6 showed that most iPS clones had a normal set of 40 chromosomes, but some clones had abnormalities, such as missing or extra chromosomes7.

These characteristics — gene expression differences and unique genetic fingerprints — ruled out the possibility that these iPS cells were accidentally contaminated with preexisting ES cells8.

These series of tests also solidified the iPS cell's identity as a novel, robust entity. But another fundamental question lingered: how did just four factors trigger such a profound cellular reset?

The architects of induced pluripotency

The earlier discovery of loosened epigenetic "locks" held a clue. One of the new additions in the “reprogramming toolkit”, c-Myc, does not directly target genes that are required in reprogramming but plays the role of a “genetic crowbar”, prying DNA from their histones and opening up the genome to make it easier for Oct3/4 and Sox2 to reach their targets. Its other role is to activate genes that drive cell growth and division, which help cells multiply during reprogramming.

The final key protein, Klf4, balances two opposing roles: activating stem cell genes and managing cell survival versus growth. Klf4 blocks p53, a protein that normally suppresses stem cell genes. Since iPS cells have lower p53 levels than regular cells, this allows these other stem cell genes to stay active, maintaining the cells’ reprogrammed state. Klf4 also slows cell division by activating p21CIP1, a protein that halts cell growth. c-Myc counteracts this by blocking p21CIP1, allowing cells to keep dividing during reprogramming. This balance between Klf4’s “brake” and c-Myc’s “gas pedal” is crucial: too much Klf4 could stall growth, while too much c-Myc might kill cells. Together, they help reprogram cells efficiently without derailing the process.

The reprogramming lottery

Despite the promising results, a key puzzle remained: why did only a fraction of treated cells actually become iPS cells? Yamanaka and Takahashi wondered whether only rare, pre-existing stem cells within skin cells (~0.067%) were being converted. Re-testing using stem-cell-rich bone marrow, however, yielded similarly low efficiency, suggesting the bottleneck wasn't the starting cell type, but the reprogramming process itself.

So, what, then, makes it such a rare event? They considered a few possibilities. Transformation might require hitting a precise "Goldilocks zone" for the four factors' activity. While the cells do have sophisticated internal controls to tightly manage their protein levels, there are limits to those controls. Additionally, reprogramming could rely partly on chance. Perhaps spontaneous genetic mutations during culture play a role, or maybe it’s the retroviruses themselves acting as genomic wildcards. Inserting their DNA cargo haphazardly across the genome (~20 times per cell!), they could accidentally disrupt important genes or activate harmful ones. This reliance on precise internal conditions and unpredictable genetic events likely explains the low efficiency, highlighting the need for more controlled methods to improve it.

Another possibility was that nature's reprogramming recipe in egg cells might be more complex than just four factors. Incredibly, two factors which are crucial in iPS cells — c-Myc and Klf4 — are not important in egg cells; in fact, c-Myc isn’t even present at all. Egg cells do, however, contain cousins of these factors — L-myc and Klf17 or Klf7 — suggesting that these might fill in for the others, like backup actors taking over (or were those the backup actors?). Egg cells are totipotent — capable of creating the entire organism as well as the organs required for its development, like the placenta — so maybe these other factors are required to potentially reprogram a cell into a totipotent cell.

More pressingly, they also examined whether the introduced factors would switch off safely after differentiation. They found that in iPS cells, epigenetic silencing could be leaky, particularly for the cancer-linked c-Myc. It underscored that harnessing this power for human treatments would require further refinement, perhaps finding alternatives to risky factors. Nevertheless, this discovery fundamentally altered the landscape, paving the way for patient-specific cells to fight disease and offering a tangible path towards the dream of personalized medicine.

A revolution in a dish

Today, these four factors that induce pluripotency — Oct3/4, Klf4, Sox2, and c-Myc — are known as Yamanaka factors. Just a year after publishing this paper on mice, in 2007, Yamanaka and his team successfully reprogrammed human skin cells into iPS cells using the same four factors. The breakthrough earned Shinya Yamanaka the Nobel Prize in Physiology or Medicine in 2012, just 6 years after the original discovery was published.

In the nearly two decades since the discovery, iPS cells have been crucial in many key advancements and applications: disease modeling, gene and cellular therapy, drug discovery and testing, and personalized and regenerative medicine. For example, in 2011, researchers from Kyoto University created iPSCs from patients with Alzheimer's disease and differentiated them into neurons, allowing them to study the disease process in a dish. Ongoing iPSC research and development is also focused on blood transfusion applications like creating red blood cells, platelets, and addressing alloimmunization risks.

Even Yamanaka, driven from medicine into research by his father's death from an untreatable disease, couldn't have foreseen the seismic impact of his work. His goal was to understand disease, hoping this might pave the way for future cures. The induced pluripotent stem cell delivered that and infinitely more — sidestepping embryo debates, creating avenues for personalized cures from a patient's own cells, and revolutionizing how we study disease and test drugs9.

This essay sought to illuminate not just the what, but the how. Science, particularly at the cutting-edge, is a demanding, iterative process. Discoveries aren’t so much stumbled upon as they are wrought. The headline “Scientists Bypass Need for Embryo” couldn't convey the years spent asking the question: “is this real?” The slew of tests wasn't just procedural; it was an essential act of validation, of creative problem-solving, and of intellectual honesty — the qualities required to answer that question with some degree of certainty. It's a story of immense effort, often invisible to those outside the lab. Yamanaka's personal motivation doubtlessly fueled the dedication required for this scientific marathon, ultimately creating a technology that heralded a future of medicine his father could only have dreamed of.

You might also like:

Asimov Press

The Making of a Gene Circuit

In the late 1990s, a young physicist named Michael Elowitz decided to “program” living cells…

Read more

a year ago · 60 likes · Niko McCarty and Nehal Udyavar

Now, I’m obsessed. I can’t get enough. I’ve read about fifteen books in the last year or so, watched countless YouTube videos, and started a bioinformatics course. And my list keeps growing. The first quarter of Somers’ post was so effective in making me consider my own disinterest-to-obsession journey – (I didn’t even read the rest until months later) – that I decided to look back and examine what caused this complete change of heart.

More than anything – nature documentaries, science shows, museum visits – it was great writing that allowed me to see the world of biology differently. My interest in biology, or rather the reversal of my disinterest in biology, began when I read The Sixth Extinction in 2016, during my second year of university. Elizabeth Kolbert’s gripping writing unveiled a completely different perspective of the subject, right alongside the scientists and researchers: driving through a Panamanian rainforest looking for golden frogs, searching a littered New Jersey creek for ammonites, scuba-diving in Castello Aragonese to inspect carbon dioxide rushing out of sea vents and in The Great Barrier Reef to look at octopi and coral reefs and blue starfish and leopard sharks and giant clams. Biology, suddenly, didn’t seem just a list of facts to memorize; it was an adventure.

I still remember how I felt after finishing her book: a strange mix of wonder and tragedy, awe and despair. That narrative structure – vivid reporting and meticulous research built on a foundation of context and history – changed how I saw science and scientists. No more dry paragraphs of definitions and explanations; every discovery had a story.

I wanted more books just like that, and luckily for me, several months later in an airport bookshop in Bangalore, I came across and picked up The Gene. I wasn’t aware of who Siddhartha Mukherjee was at the time (possibly the mention of Pulitzer Prize winner on the cover influenced me), and I had no prior interest in genetics, but that book would end up completely changing my worldview on biology and non-fiction writing. If Kolbert made a crack in the dam I had built around biology, Mukherjee would go on to smash the whole thing down to pieces.

One of the stories in the book, the discovery of the gene that caused Huntington’s disease, moved me tremendously when I first read it a few years ago. It’s the perfect example of the amount of effort that goes into a scientific discovery that then ends up as a single sentence in a textbook; in this case, that Huntington’s disease is a hereditary, neurodegenerative disorder caused by a mutation in a single gene.

The story of finding that mutation would make a thrilling movie: a young woman named Nancy Wexler, devastated by the news that her mother has been diagnosed with Huntignton’s and that she and her sister would have a 50-50 chance of getting it, decides to devote her life to solving this medical mystery. Her quest takes her from nursing homes in Los Angeles to interdisciplinary scientific workshops in Boston to stilt villages surrounding Lake Maracaibo in Venezuela. Her decade-long blood and skin sample collection efforts there would create the largest family tree with Huntington’s, leading to the first genetic test for the disease, followed by locating the precise genetic mutation that caused it1.

The gene sequence had a strange repeating structure, CAGCAGCAG… continuing for 17 repeats on average (ranging between 10 to 35 normally), encoding a huge protein that’s found in neurons and testicular tissue (its exact function is still not well understood). The mutation that causes HD increases the number of repeats to more than forty – a “molecular stutter” – creating a longer huntingtin protein, which is believed to form abnormally sized clumps when enzymes in neural cells cut it. The more repeats there are, the sooner the symptoms occur and the higher the severity2.

Nancy herself opted not to take the genetic test she helped create. “If the test showed I have the gene,” she wrote in 1991, “would I continue to feel the happiness, the passion, the occasional ecstasy I feel now? Is the chance of release from Huntington’s worth the risk of losing joy?”. In 2020, at the age of 74, she revealed that she had Huntington’s. The public acknowledgment was not a surprise for those close to her – for the last decade, they noticed her gait slowly deteriorate, speech slur, and limbs jerk in random directions, the same characteristics she saw in her mother half a century ago, and in the hundreds of Venezuelan patients she tended to ever since.

There’s still no cure for Huntington’s disease, but every time I hear about progress on cures, I feel a rush of emotions, like I have a personal stake in its invention. I really wish to see one found within Nancy Wexler’s lifetime; this movie deserves a happy ending.

Pick a field in biology, or a slice of history, and you’ll find countless stories just like this. Mischievous Watson and Crick figuring out the structure of DNA after getting a peek at Rosalind Franklin’s crisp x-ray crystallography photograph; Baruch Blumberg discovering hepatitis B after locating the antigen in the blood of an Australian Aboriginal, and beating NIH to its cure, the world’s first cancer vaccine; James Simpson systematically inhaling various vapors and recording its effects in the search for a better anesthetic, resulting in the discovery of chloroform; Andreas Vesalius taking prisoners’ corpses hanging in the gallows in 16th century Paris and, along with painter Andrea Mategna, publishing nearly 700 incredibly detailed drawings of the human anatomy.

History and stories may not be immediately applicable, but when used as a key ingredient it makes the discoveries more majestic, more impactful. That’s what I love about Mukherjee’s writing: it’s a unique stew of history, biography, experimental methods and results, scientific findings and their significance, seasoned well with personal anecdotes, and presented with the candor of a physician and the artistry of a poet. The context creates a kind of multiplier when the mind-shattering discoveries are explained – how a genotype gives rise to a phenotype, how cancer works, how a heart beats or a bone mends itself or a brain remembers a memory. Like the climax of a movie scene, the beauty and immensity of the discovery or the invention feels far more compelling after following the steps that got us there.

Every discovery might not have an entertaining backstory, but even when focusing on just the phenomenon, great technical writing has this striking ability to make you see the world differently. The same molecule or cell or organ, theory or experiment or discovery, suddenly seems monumental, like it’s the most important thing in the world. It makes you think: why didn’t I learn about this before?

One of my favourites is the way Mukherjee describes how a neuron communicates in The Song of the Cell:

Imagine the nerve, first, in its “resting” state. At rest, the internal milieu of the neuron contains a high concentration of potassium ions and a minimal concentration of sodium ions. This exclusion of sodium from the neuron’s interior is critical; we might imagine these sodium ions as a throng outside the citadel, locked out of the castle’s walls and banging at the gates to get inside. Natural chemical equilibrium would drive the influx of sodium into the neuron. In its resting state, the cell actively excludes sodium from entry, using energy to drive the ions out…

[...] The dendrites are the site within the neuron where the “input” of the signal originates. When a stimulus—typically a chemical called a “neurotransmitter”—arrives at one of the dendrites, it binds to a cognate receptor on the membrane. And it is at this point that the cascade of nerve conduction begins.

The binding of the chemical to the receptor causes channels in the membrane to open. The citadel’s gates are thrown ajar, and sodium floods into the cell. As more ions swarm in, the neuron’s net charge changes: every influx of ions generates a small positive pulse. And as more and more transmitters bind, and more such channels open, the pulse increases in amplitude. A cumulative charge courses through the cell body.

The mental picture of a throng of sodium ions locked out of the castle walls is so helpful and convincing. I can see, in my mind’s eye, these shadowy ions banging at the gates to get inside, like an invading army. Then, after the neurotransmitter binds to the cognate receptor, the sodium ions don’t just enter, they flood and swarm in; the membrane doesn’t just open, its gates are thrown ajar. The metaphor makes the chemical process relatable without leaving out the details; the vivid language romanticizes it, creating a mental picture that not only stays with you, but makes you want to learn more.

A little later in the chapter, Mukherjee writes about neural connection in the fetus:

Neural connections between the eyes and the brain are formed long before birth, establishing the wiring and the circuitry that allow a child to begin visualizing the world the minute she emerges from the womb. Long before the eyelids open, during the early development of the visual system, waves of spontaneous activity ripple from the retina to the brain, like dancers practicing their moves before a performance… This fetal warm-up act—the soldering of neural connections before the eyes actually function—is crucial to the performance of the visual system. The world has to be dreamed before it is seen.

There’s something about this evocative language that leaves a sweet, lingering imprint on my mind — a new set of neural connections; my own throng of sodium ions banging at the gates, my own ripples. The details – which ions, the name of the receptor – might get murky after the passage of time, but the sweet feeling remains, like a memory of a heavenly meal; you may have forgotten the exact taste, but the feeling of satisfaction lingers, and occasionally, when it enters front and center, you might imagine visiting the restaurant (or home) once more.

That’s what I feel after reading books like this – the belief that I’ll revisit it, relive it, relearn it. It fills up a reservoir of curiosity, and every subsequent piece of stimulus – a neurology article or academic paper shared on Twitter, a documentary or YouTube video, another book (even textbooks) – opens the floodgates, and makes you want to explore a little more. I might not have the equipment to see this cell myself, but when written like this, this world too can be dreamed before it is seen.

The more you explore, the more astonishing it gets. Suddenly, you’re surrounded by these facts that stop you in your tracks. Like the fact that there are 20-30 trillion red blood cells in our body, making up roughly 84% of all our cells, and 1.2 million are created in our bone marrow every second. Or the fact that our visual system is predictive, calculating where to move the hand to catch a ball before your visual system has fully registered its trajectory3.

One of my favorite ‘sentences that stopped me in my tracks’ comes from Nick Lane’s book, The Vital Question. He starts with carefully explaining that all cells derive their energy from a single type of chemical reaction, the redox reaction, where electrons are transferred from one molecule to another. Rust is a redox reaction: iron donates electrons to oxygen, being oxidized in the process. Same with fire: oxygen (O₂) is reduced to water after receiving two electrons (O^2-) and then two protons (H₂O), balancing the charges, and releasing heat in the process. Respiration — the process that turns our food into energy — does exactly this as well, except that it conserves some of the energy in the form of a molecule called adenosine triphosphate (ATP). Think of ATP as an energy currency, able to be stored or converted back into energy by splitting the molecule into ADP (adenosine diphosphate) and P_i (phospate). And so, he writes, “in the end respiration and burning are equivalent; the slight delay in the middle is what we know as life.”

Wait, what? The slight delay in the middle is what we know as life? I think when I first read that I might have skipped a heartbeat. I learned about mitochondria and ATP and redox reactions and aerobic respiration in high school, but I never pictured it as millions of molecular fires that keep us alive. Actually, not a million; it’s at least a quadrillion – per second.

ATP is synthesized by the fabled mitochondria, but that’s not all they do. They also regulate metabolism, participate in cell growth and death, manage calcium levels, and are involved in detoxification, hormone production, and cellular signalling. They even have their own genetic code. In fact, your mitochondria come from your mother and your mother only; they’re not genetically recombined like the rest of you. They’re remarkably fascinating; even the universally memed “powerhouse” doesn’t quite cover its capabilities.

All of this is still merely scratching the surface of wonder. I’ve only really described three examples in biology, all of which relate to human cells. But we’re just one of the millions of organisms on this planet. Bacteria, plants, fungi, insects, birds, reptiles, mammals, and everything in between, are all made up of cells. And every level – ecological, species, organism, tissue, cellular, organelle, protein, genome – has its own stories, each its own magic.

In his blog post, Somers advised to learn in small, deep slices. But I took a different approach: I went shallow and wide. Kolbert, Mukherjee, and Lane inspired exploring adjacent domains, and so I read about epidemiology, drug discovery, gene editing, molecular biology, systems and synthetic biology, immunotherapy, and memoirs from surgeons, cancer patients, and “biology watchers”. Even my fiction choices started to exhibit a biology tinge: The Shell Collector, The Covenant of Water, The Overstory. Eventually, I started seeing biology everywhere — the roots of a sidewalk tree battling with concrete, a group of sparrows frolicking in a bush, a young woman in an air cast fiddling with her crutches — as if it escaped the pages and began whispering its presence wherever I went.

Last summer, I went scuba diving for the first time in my life. I’ve wanted to go since I was a teen, a desire amplified after reading Kolbert’s adventures and watching ocean documentaries. After years and years of postponing, I finally pulled the trigger and flew to Puerto Vallarta to get Open Water certified. I could fill an entire essay with just this certification experience — the anxiety-inducing pre-dive coursework that essentially just lists the many ways you can get seriously injured or die; the silly awkwardness of training in a Mexican hotel pool surrounded by curious onlookers; the ear injury I sustained after my first ocean dive, where a rupture caused by improper depressurization caused middle ear fluid to flood my right ear canal, leaving me with partial hearing loss for a week (even PADI’s intimidating coursework could only do so much) — but I will focus on just the experience of my second dive here.

It was a picture-perfect day in Puerto Vallarta: deep blue skies, fluffy cotton-candy clouds floating above, a momentary cool breeze tempering the unrelenting summer humidity. As our boat sped along to Playa Majahuitas, about a 40 minutes ride from the main pier, I watched the lush green hills roll by just behind the shore, the ocean shimmering as the sun flung silver disks across its surface. During the ride, I asked the couple sharing the boat about their scuba experiences, and, again, I got a common response I still couldn’t relate to: that it was meditative — it was where your problems of land disappear, and you get to be a visitor in the home of sea-life, a polite guest just observing.

Our dive spot looked like a painting: water so clear you could see schools of fish just by peering over the edge of the boat. Just before we began, we got a surprise visit from a manta ray – this enormous, ethereal creature silently gliding under the water, just flicking the tips of its wings above the surface, as if to say hello, and welcome us into its home.

After the dive that day, I understood what the couple meant. I felt a lot more comfortable with my equipment second time around, and so, no longer apprehensive about buoyancy or breathing rate or how deep I was, I finally felt free to fully take in my surroundings. I fell into a gentle rhythm: inhale, listen to the hiss of the regulator, exhale, watch the bubbles float away. You’re distinctly aware of each and every moment, mind blank and in awe of the world around you: a large school of Cortez wrasses passing by; a camouflaged octopus hiding under the seabed; a moray eel sticking its neck out of a little hole, an angry look on its face, as if you’ve just disturbed its sleep; the vast, splendid diversity of corals – you can see it living, with little, wavy hand-like appendages collecting bits of floating food to eat, with tiny fish swimming in and out and around, as if playing a game of tag.

It was truly marvelous. Colors, too, are more vibrant underwater, as if the gods enhanced saturation as a gift to those that dare venture below. The body of spotted boxfish are a glittery blue, and the yellow speckled top shines in contrast. The corals too are rich: deep oranges, yellows, greens and browns. Even ocean documentaries, with their film-grade color editing, don’t capture the true shades.

During the boat ride back, I had this incredibly calming bliss completely take over my body. (Maybe that’s also what people attribute to its meditative quality, although meditative isn’t exactly the right word). For me, the whole experience would mark the start of a gradual realization that I wanted my role in biology to be more than just reading. My favorite science writers – Kolbert, Mukherjee, Lane, Lewis Thomas, Donald Kirsch – all wrote from experience, and if I wanted to write, or create, like that, I’d have to experience the world too. I began piecing together the things that had been swimming in my mind: namely, how to combine my past passion, interactive learning, with my latest obsession, biology.

I have since restarted working on my website, Newt Interactive, to make interactive articles and accessible simulators for topics in biology. I too, like Somers mentions at the end of his blog post, want to bring the three dimensional nature of biology to life. The subject is teeming with fascinating phenomena that remain hidden or inaccessible to those outside scientific and research communities. Occasionally, I’ll come across something incredible — like a video of a molecular motor in action — but the sheer marvel of that just fundamentally doesn’t click unless you’re already well versed in the subject4.

I hope to bridge this gap and make some of biology's intricate mechanisms comprehensible and awe-inspiring for everyone. I’ve started with an interactive series on systems biology (and wrote about my idea and motivation behind it in a previous post), as well as some standalone simulators for a few concepts: coherent type-1 feed forward loops and genetic circuit evolution, for two. My goal is to work my way up to more sophisticated simulations, tools, and interactive articles that will help illustrate, and importantly, allow you to play with, more advanced concepts. In addition, I’d like to generally write and draw more as well (also started with this by making my first science graphic and biological math model for Asimov Press).

Stories of science can elicit all kinds of emotions: joy, sadness, enchantment, heartbreak, optimism, valiance, apprehension, intrigue. I find, however, that one theme seems to be consistent among the characters: curiosity. This shouldn’t come as a surprise, of course, but what I hadn’t anticipated was how infectious it could be. Just reading about these scientists — their history, theories, efforts, mistakes and unwavering dedication to truth — kindled an active curiosity in me. I don’t think I have the patience to do what the scientists I read about did, experimenting day after day, week and week, year after year, exploring a small sliver in the “infinite vastness of biology”. And, since my curiosity started and ended with books, I didn’t think there was a meaningful role I could play. I couldn’t hear the calling.

But now I’m not so sure. I have this recurring desire to look down a microscope, and see a cell live its life, see its components swimming, squirming, dividing. I want to see a sequencing machine take in an organism’s DNA and spit out all its nucleotide bases; to hold a test-tube with genetic material that I edited with CRISPR-Cas9; to roam around a laboratory and peek at each bench’s weird collection of tools and equipment and liquids, slide my feet across the polished laboratory floor, smell the lingering scent of disinfectant; to go on more dives and hikes and explore the breathtaking diversity of life. It’s not quite a calling, more like hearing a faint ringtone in a distant room. You’re not sure if your phone’s ringing or your mind’s making the sound up. Maybe this time it’s worth taking a look.

In my previous post, I wrote about the simulator I built to explain how genetic circuits evolve through random mutations. This time, let’s explore a specific type of genetic circuit — the coherent type-1 feed-forward loop — which I first learned about in Chapter 3 of An Introduction to Systems Biology.

Feed-forward loops (FFLs) are among the most common network motifs in biology — recurring patterns that show up far more often than you’d expect by chance. In a feed-forward loop, a transcription factor X regulates a second transcription factor Y, both of which regulate gene Z. So there are two paths to activation for Z, a direct path (X) and a delayed path through an intermediate (Y).

There are 8 types of FFLs, which can be split into two groups: coherent and incoherent. In coherent FFLs the indirect path (X → Y → Z) has the same sign (positive or negative) as the direct path (X → Z). In incoherent FFLs, the signs are opposite. The diagram below shows all the combinations:

The specific one I built a simulator for is the coherent type-1 feed-forward loop (C1-FFL), where both paths are activators. You can think of this like so: imagine you're at a company with a particularly bureaucratic approval process. To achieve your goal of geting a project approved by the VP (Z), you need sign-off from both your direct manager (X) and the department head (Y). Once you submit your project (the signal), your manager approves really quickly, but the department head is busy and takes longer to approve, before reaching the VP’s desk.

This is exactly how a C1-FFL works in biology. When a signal (Sx) comes in, it activates both a fast-responding transcription factor (X), which in turn activates the intermediate transcription factor (Y), both of which regulate the target gene (Z). If you’re familiar with some programming, this essentially functions as a biological AND gate. This architecture encodes some interesting behaviours, and I built a simulator to play with and explore exactly this. I encourage you to check it out before reading the details of the behaviours below.

Circuit Behavior

In the chapter, Uri Alon describes the two behaviours this type of circuit generates: a sign-sensitive delay and persistence detection. You can visualize these through the simulation. Sign-sensitive delay refers to the fact that starting the production of Z has a delay (shown by yellow boxes in the screenshot below) because it depends on both X and Y, but stopping production, after either of the two transcription factors is removed, does not. This delay works as a persistence detector for ON pulses: Z only responds if there is a persistent signal from X AND Y; or, in other words, it is filters out noise in the system.

The Arabinose system in E.Coli

A great example of C1-FFL in action is the Arabinose system of E.Coli. Alon explains the whole thing in a lot more experimental detail in Chapter 3, so I encourage you to check it out there, but I’ll briefly summarize it here.

Arabinose is a sugar, but it’s only used by most cells if glucose is not present as glucose is a far better energy source and arabinose is costlier to digest. So, E.Coli only wants to build the machinery to digest arabinose if glucose is not present and arabinose is. The absence of glucose is represented by a small molecule called cAMP, which is the signal that activates the transcription factor CRP (let’s call this X from our example). Another transcription factor, AraC, senses arabinose (this can be Y). These two regulators are part of the system to produce arabinose utilization proteins araBAD and araFGH.

Experiments show that there’s a delay in producing the final proteins compared to simple regulation when both signals are ON, but no delay when stopping protein production. This is incredibly useful when there’s a huge cost associated with building particular proteins for a cell, which in this case is the arabinose utilization proteins. This circuit design enables E.Coli to only produce them when they’re absolutely certain that they are necessary (persistence detection), and immediately stop expensive production as soon as a better option is available (in this case, the appearance of glucose and so disappearance of cAMP.

Future Ideas

I hope that this simulator allows you to play with this incredibly cool facet of life, and try different parameters and see what happens. It’s the first in this specific kind of interactive circuit simulation I’ve always wanted to make. Every time I learn about a concept like this, I wish that I could visualize it in real-time, change some parameters, and see what happens. They’re also fun to make and I learn a lot in the process.

I can also use this as the foundation to build similar simulators for other circuit designs with other features, like incoherent type-1 FFLs (speed up ON and biphasic response), single-input modules (temporal expression, just-in-time production), positive autoregulation (bistability), and many more.

It also sets up a future interactive I would like to build, where you can drag-and-drop to create and run your own genetic circuit. This step brings that goal a little closer.

If you find this interesting as well, or have any feedback/questions or find any issues with the simulator, I’d love to hear from you! Otherwise, thanks for checking this out, and I’ll see you in the new year!

In Part Three of An Introduction to Systems Biology, Uri Alon introduces the concept of modularity in biological networks. A modular system is one that can be broken down into smaller, independent components that each have a specific function. The simplest analogy is LEGO: large, elaborate structures can be built from a small group of blocks. If some of your blocks go missing — accidentally vacuumed and thrown out, lost to another dimension after rolling under the couch — you can simply get identical ones, and continue building with doing a lot of re-working. Another example is cars: batteries, transmissions, motors, tires, etc. can be easily replaced without taking the entire vehicle apart.

Organisms tend to show similar modularity in their structures at all levels. At the organ level, Alon provides the example of our lung and liver. The liver supplies glucose and blood proteins to the body, the lung supplies oxygen and removes CO2; the liver needs the lung’s oxygen to function, the lung needs the liver’s glucose to function. You could, theoretically, replace the lung with something that matches its output, and the liver will continue to function (like with a heart-lung machine temporarily, for example). They’re modules in the larger system, our body.

A similar structure is also found in the cellular level: each cell has subparts, like the nucleus, ribosome, etc. that have their own function. And it’s also found at the level of proteins, with sub-structures like different binding and dimerization sites.

LEGOs and cars are engineered to be that way: modularity is beneficial for cost-saving and mass-manufacturing, for instance. Biology. however, didn’t have an MBA graduate and factory engineer pore over best practices. So, why did it evolve modularity as well?

To begin answering this question, Alon described some results of computer simulations. He reported that, in simple computer simulations of evolution, non-modularity is the norm. He writes:

In evolutionary simulations, a population of networks is evolved by randomly adding, removing and changing connections between nodes – and even duplicating and recombining parts of the networks – until the networks perform a given computation goal, that is, until the networks give the correct output-to-input relationship.

He specifically goes on to describe a network of NAND gates, and mutations that changed the network. I thought this sounded really interesting, so I decided to create a simple computer simulation that does some of this and see how a network “evolves” myself. You can check it out on my website, Newt Interactive! I recommend playing around with it first and getting a feel for what’s going on and noticing some patterns.

This simulator comes in two modes: Mutation and Generation. In Mutation, you simulate a random change to the circuit — in this case, it’s just changing the connections. In Generation mode, N variants of the circuit are simulated, each with a single mutation, and from these variants (and the original) the highest scoring circuit is selected. This is a rough mimicry of natural selection, and shows different behavior compared to just random mutations.

If you try out the Generation simulation, you’ll see the same phenomenon that Alon goes on to discuss in the chapter: logarithmic slowdown. Essentially, the circuit quickly improves at first, and then hits a wall and stays at the same spot for a long time. (Mine’s capped to 100 generations for now, but usually getting to the perfect score takes around 10,000 generations).

Why make this

I wanted to build this for a few reasons:

1. as I was reading the chapter, I wanted to visually see the changes per mutation/generation (and what happens when I change parameters)

2. the process of building something usually brings a deeper understanding of the problem, as well as coming to the same insights but through first-principles

3. it’s a stepping-stone to build more complex simulations, like the one required to see modularity evolve, which I’ll explore in a future Part Two post.

Modularity and Simulator 2.0

The reason these simulations evolve into non-modularity is because it’s the most efficient, but often networks get stuck in local maxima that delay reaching that perfect solution — the logarithmic slowdown. As Alon reveals later in the chapter, modularity is achieved by evolution when the goal varies periodically, as long as both those goals tend towards the same direction; in other words, two goals where working on one gets you closer to the other. In short, this change provides an opportunity for circuits to exit local maxima without veering too far from either goal.

In simulations, this is done with circuits and varying logic goals, and the results are explored in detail at the end of the chapter in the book. Modular solutions are less efficient (in this case, the modular one requires 11 nodes, versus the 10 for the non-modular one), but reach their goal faster.

My goal here is to make this part of the simulation as well. Like with the above version, I’d like to see the changes as they happen and watch “evolution” reach modularity. This requires some upgrades to the circuit simulation I have, which I’ll roughly list here, but will build in the future and explore more deeply in Part Two:

• mutations should include ability to add/remove nodes (right now only changes connections)

• more variations created per generation (right now capped to 20)

• up to 10,000 generations (right now capped to 100)

  • ability to skip ahead some number of generations (having to click 10,000 times would be pretty painful)

• switching circuit goals (either manually, or automatically for modularly varying goals)

• automatic visual reorganizing of circuits (right now it can get quite messy, although I’m not quite sure how to do this)

• measuring modularity (there are algorithms that do this)

• other visual indicators, like the mountain climbing visual in the book?

Of course, as I build it, other ideas will come forth or some of these might not seem that great (or impractical). But should be interesting.

If you find this interesting as well, or have any feedback/questions or find any issues with the simulator, I’d love to hear from you!

As you follow along, you’ll learn the concept of network motifs through the lens of gene regulation in E. coli. We start with a seemingly chaotic network of hundreds of genes and their interactions, but then zoom in on a fascinating pattern: genes that regulate themselves — autoregulation.

What makes this particularly interesting is how we determine whether these patterns are meaningful or just random occurrences. Through interactive visualizations, I show how to:

• Compare "real" biological networks with random ones

• Generate and explore different random network configurations

• See how network statistics change as parameters are adjusted

I also break down some of the math a little more clearly, like explaining where the standard deviation equation for a network comes from (the side drawer with extra info in the screen recording). It’s what’s used to compare whether a pattern is a motif — i.e. whether it occurs far more often than it would at random. In the article, you’ll learn how in a portion of E.Coli, autoregulation occurs at a frequency about 35 standard deviations away from random.

This is the first in the chapter about autoregulation in biological systems. The next part will explore why negative autoregulation (genes that repress their own expression) is so prevalent in nature and what advantages it provides to organisms.

Thanks for reading and stay tuned!

(If you’re wondering why this is #4, articles 1-3 were shared in my introductory post on my Systems Biology series). You can also check them out here:

1. Transcription Network Basics: the fundamental concepts of transcription factors and gene expression

2. Activators and Repressors: the two types of transcription factors and their mathematical models (Hill function)

3. Dynamics and Response Time: how cells respond to changes in signals and modelling how long that takes

While reading the Chapter 2 in An Introduction to Systems Biology, I learned about the concept of network motifs, which is a pattern in a network that occurs much more often than would at random. In transportation networks, a network motif could be the emergence of hubs — single locations that connect to and from a lot of other locations. In biological transcription networks, a network motif could be the emergence of master regulatory transcription factors — single proteins that play a role in regulating a lot of other genes. Or it could be the discovery of autoregulation — transcription factors that regulate themselves — which is what the chapter in the book explores.

Regardless of the application, in order to know that something occurs much more than at random, it needs to be compared to the random. Given that we have a real network, how can we compare it to a random one?

A simple model of random networks was created by Hungarian mathematicians Paul Erdős and Alfréd Rényi, called the Erdős-Rényi (ER) model. The model comes in two variants: G(n,p) and G(n,M).

• With G(n,p), you start with n number of nodes and then connect each pair of nodes (create an edge between them) with a probability p, independent of all other pairs. If the probability is 1, all nodes will be connected. If it’s 0, none will be connected. If it’s 0.5, then whether a connection is made is like tossing a coin.

• With G(n,M), you start with n number of nodes and randomly select M edges (from all possible edges) to connect the nodes. This is a useful method for generating graphs when you want a specific number of edges.

The generator I built is for the G(n,M) variant. If you liked it and would like to play with the G(n,p) one as well, or have any other questions, please comment or reach out to me!

Building this model from scratch taught me a lot of details about how it works; the process made me ask a lot of questions about what steps were right, details that I did not consider after just reading about it. The graph generator is a stepping stone in working on my interactive Systems Biology series. In an upcoming article on autoregulation, you will see how real networks are compared to random ones, and how by using this comparison we can find that autoregulation is a network motif in biological transcription networks, and what advantages that might hold for organisms. If you’re interested in this topic, you can start with Part One: Transcription Network Basics, or check out the whole series (so far)!

Over the last few months I've been trying to find my own creative way of sharing what I've been learning in biology, and this feels like the right combination of my early passion — building interactive learning content — and my most recent obsession — biology.

I’m beginning with systems biology because it’s an ideal subject for interactive learning. It deals with intricate networks of genes, proteins, and cellular processes that can be challenging to visualize and understand through text alone. I want to bring systems biology concepts to life by incorporating interactive elements, allowing you to manipulate variables and see the results in real-time. Here’s a simple example exploring the Hill function for an activator:

So far, I’ve written three articles as part of an Introduction to Transcription Networks:

1. Transcription Network Basics: the fundamental concepts of transcription factors and gene expression

2. Activators and Repressors: the two types of transcription factors and their mathematical models (Hill function)

3. Dynamics and Response Time: how cells respond to changes in signals and modelling how long that takes

The articles and the series follows along Uri Alon’s textbook, An Introduction to Systems Biology. I’ve been mesmerized by the book and wanted to play around with some of the concepts introduced and also deepen my understanding (especially some of the math), which inspired building this series.

Apart from the interactive chart shown above and the main text of the articles, here are a few features you’ll find in my series:

• fun diagrams:

• detailed math derivations and explanations:

• and yes, more kinds of interactive charts and diagrams:

What’s Coming Up

Several months ago, I learned about FoldIt, a game developed by David Baker’s team at the University of Washington that lets players fold a string of amino acids into protein structures, and get in-game feedback as to whether the folding was correct or not. I downloaded the game and gave it a shot — it was originally released in 2008 and, even with the newer versions, feels a little old now, but it still does a fantastic job of allowing you to “feel” protein folding.

I also learned that Demis Hassabis, the co-founder of DeepMind, played FoldIt in his undergraduate days. In fact, FoldIt was one of the inspirations to decide to use the same technology that was beating Lee Sedol at Go — AlphaGo — to take a crack at protein folding — now known as AlphaFold. (Baker and Hassabis, as you might have heard, were recently awarded the 2024 Nobel Prize for Chemistry for computational protein design, along with John Jumper).

Learning about FoldIt and Hassabis’s role in it was another reminder for me about the importance of games, and how, when done correctly, can have huge long-term impacts. They can be great introductions to complex topics, where players and learners can develop intuition about a topic or problem before having to dive into the, usually, intricate technical details.

These introductory articles are a stepping stone towards creating those complex interactives and educational games in the future. As we go deeper in the series, I want to make more elaborate interactives that visually explain more complicated topics like coherent and incoherent feedforward loops, single input modules (SIMs), and how specific genetic circuitry can enable stability, memory, and oscillation — topics that are the building blocks to understanding relatable experiences like pain, heart beats, and circadian rhythms. I want to work my way up to more sophisticated simulations and tools that will help illustrate more advanced concepts in systems biology and beyond. I want, essentially, to build my own FoldIt.

I believe that this approach – combining clear explanations with hands-on, interactive elements – will make learning about systems biology both more accessible and more fun. Whether you're a student, a professional in a related field, or simply curious about how cells work, I hope you'll find this series informative and enjoyable.

Please subscribe on Newt Interactive if you’re interested in following along (there’s a box at the bottom of the page). I will post regular updates about it here, as well as continue to use this space to write and share pieces that are not interactive-based. If you check it out or find this topic or format interesting, I would love your feedback. I’m also happy to answer any questions or generally just chat!

Additional Links

• FoldIt: you can find the game here

• How AI Revolutionized Protein Science, but Didn’t End It: A brilliant article by Yasemin Saplakoglu explaining the history of discovering the shape of proteins and how they fold, how that led to advanced computational tools like Rosetta and AlphaFold, and a breakdown of what AIs “solving” protein folding really means. This was where I learned about FoldIt, among many other things. I wrote about it in July but sadly it never left my drafts.

1. ESM3: Simulating 500 million years of evolution with a language model

by EvolutionaryScale Team

EvolutionaryScale came out of stealth to announce a new LLM for biology, with the aim to use “AI to engineer biology from first principles.” They also announced that they’ve managed to use this model to create (or is it discover?) a new green fluorescent protein (GFP) — the kinds that allow jellyfish and corals to glow — which they’ve named esmGFP. The AI-generated protein is “only 58% similar to the closest known fluorescent protein”, a change they claim is estimated to be “equivalent to simulating over 500 million years of evolution.”

That’s… a lot to process. I’m not an expert on LLMs, but I do know the basics of how it functions, so there were some interesting aspects they mention, along with some questions (I assume some will be explained in more detail in the preview paper, which I’ve not read yet):

• they’ve transformed the three fundamental properties of biology — sequence, structure, and function — into a new token data format (set of discrete alphabets) so that the model can be trained on all simultaneously and promise “unlocking emergent generative capability.” I’m curious about this data transformation; I think it’s important that the connection between these three properties are promoted in training, versus just depending on a single data point, like sequence, for example.

• they’ve augmented the training dataset with “hundreds of millions of synthetic data points” because they have limited amounts of annotated experimental data. I wonder if the limitation is experimental data, or experimental data annotations here. If it’s just annotations, how can more experimental data be annotated at scale? Or is it not required as much because synthetic data is quite a capable substitute? And how is the validity of synthetic data tested prior to training?

• ESM3 is self-improving, but they also mention that “feedback from laboratory experiments or existing experimental data could also be applied to align ESM3’s generations with biological success.” How would this be done in practice, taking into account the data transformations? What does fine-tuning with this model look like?

• they also mentioned that ESM3 generated protein candidates with a chain-of-thought process, after being “prompted with the structure of a few residues in the core of natural GFP.” Can this chain-of-thought be visualized? If the claim is “equivalent to simulating 500 million years of evolution”, looking at how the model got to where it ended up on — and what the intermediaries looked like — could be interesting, and also maybe provide insight into the models “evolutionary path” for the engineered GFP.

I am, however, a bit skeptical about the responsible development and public benefit company status. I think that things can change, and a single lucrative discovery might lead to changes in company structure or slight changes in the accepted community values, guiding principles, and commitments, like with OpenAI. That being said, I’m excited to read what scientists and researchers have managed to do with the model so far, and what they’ll use if for in the future, as well as to learn more about AI and biology.

2. The Crimes Behind the Seafood You Eat

by Ian Urbina

This is a difficult read, but an important one. Urbina’s reporting is stellar, following sources and leads for years, and going to extreme lengths to give a voice to the unheard. (One I particularly admire was how his team, when denied access to come onboard a vessel, followed it in a skiff and threw bottles filled with a paper with questions and rice for weight on-board, which were then answered, put back in the bottle, and thrown back to them).

The story of Aritonang is tragic, and to think that he’s just one of tens of thousands — millions if you take into account other industries — of people trapped in modern slavery. Life on a distant-water fishing ship is arduous in the best of conditions, so it’s impossible to grasp how terrible and hopeless life must feel when servitude is tacked on top.

It’s difficult to conceive of a near-term solution to any of this. Like Urbina explains, distant-water fishing fleets are extremely difficult to track, and ensuring vessels meet international maritime standards is nearly impossible to ascertain, especially when accounting for geopolitical alliances. Import bans are difficult to uphold because seafood from the illegal vessels can be secretly transferred to other vessels, essentially white-washing it. After it reaches a processing facility near the port, it again gets harder to separate the legal from the illegal, and so, likely a significant portion of important seafood — particularly squid, that this report details — comes from illegal vessels.

In between processing sorrow for Aritonang and others like him, my mind kept considering one possible long-term solution to both this crime and the problem of overfishing: lab-grown meat. I remember, while I was doing my Terra.do class last year, we had discussed lab-grown meat as a future alternative for meeting worldwide demand while still reducing meat’s climate footprint (land for cattle, overfishing, etc.). Companies like Wildtype Foods, for example, already grow salmon from cell culture in their microbrewery-like lab, and as of October 2021, produced 50,000 lbs of it per year this way. I’d guess there’s no reason squid can’t be made the same way, or beef or chicken. The question is: can it scale? And, if so, can the public be convinced it’s the same?

3. Anthony Bourdain’s Moveable Feast

by Patrick Radden Keefe

I’m a huge Anthony Bourdain fan. I, like millions around the world, watched Parts Unknown and No Reservations with deep interest, intrigued by this brash-yet-likeable man’s incredible ability to combine discovering and devouring mouth-watering local cuisine with the region’s cultural and political discourse. I read Kitchen Confidential almost two years ago, and marvelled at both his wild upbringing and the cocksure, gruelling kitchen environment, written in much the same vigor and punchiness.

I’m also a huge fan of Patrick Radden Keefe. I first came across his writing through his book, Say Nothing, the shocking and violent account of the Troubles and its impact on Northern Ireland. I read it three-and-a-half years ago, and I can still picture some of its haunting imagery of countryside murders and terrorist bombings. I naturally kept track of Keefe’s writing since, so when Empire of Pain launched I immediately added it to my list. I finished it last month, and again, I was in awe at Keefe’s ability to report, so clearly and deftly, on a topic that for decades had been purposely twisted and concealed in a shadow of deceit and mystery.

As I looked through Keefe’s older New Yorker articles, I was surprised to find that he had published a short essay on Anthony Bourdain the day after Bourdain died, which mentioned a profile (this one) he had written a couple of years earlier. It felt like two of my worlds colliding. I had no idea that Keefe, traveling with Bourdain for the profile, met up with him in a Hanoi bar the day after that famous dinner with Obama at the bún chả restaurant was filmed. Reading this felt like witnessing behind-the-scenes footage from that popular segment, filmed by a character I had assumed distinct from the world of Bourdain. (I had, in fact, visited (but didn’t eat at) that same restaurant early last year when I went to Hanoi, vaguely following the trail made by his show.)

Crossover-episode-style fun aside, this article is another literary gift from Keefe. He, as I also felt in Say Nothing and Empire of Pain, seems to be a master in probably what is the most important skill as a writer: getting the reader to read the next sentence. Maybe partly it’s the thrilling subjects he writes about — violence and murder in Northern Ireland, nationwide pharmacological white-collar crime — but nonetheless I find myself gliding through the text, completely engaged, almost incapable of giving up attention in my otherwise occasionally-distracted mode of reading. Even if you’re somehow not a watcher of Bourdain or a reader of Keefe, this is a pretty good place to start.

4. How 3M Executives Convinced a Scientist the Forever Chemicals She Found in Human Blood Were Safe

by Sharon Lerner

This was another frustrating read about scientists and executives actively concealing important information from the public, and then, once the crimes comes to light, getting away completely unscathed. After last week’s carcinogens article, this one sheds light on a new threat: forever chemicals. Despite 3M knowing, since the 70s, that their fluorochemicals were seemingly everywhere and in everyone, they hid that information from employees, state and federal authorities, and the general public, until they were forced their hand. Lerner follows the story of Kris Hansen, and her conflicted role not in hiding, but rather ignoring what she knew to be a very serious problem.

This article also serves a great example of why having a company police itself for environmental and public health matters is unlikely to work; the incentives are just not there for executives to publicize their mistakes early, when the damage is still containable. Instead, you end up with this tragicomic if-I-didn’t-look-it-didn’t-happen story, where each person in the hierarchy wonders why no one is believing and following up on this very serious issue and eventually ignores it. Years or decades later, when the people in the story look back at their actions, the best you can get is a feeling of remorse, like with Hansen, and at worst malice, like with her boss, Johnson.

Another pattern I keep finding is of companies settling lawsuits with the inclusion that they don’t admit fault or liability. 3M paid $850 million with this condition to settle the lawsuit brought by the Minnesota attorney general. It was also the modus operandi of Purdue Pharmaceuticals and the Sackler family, as described in The Empire of Pain. The fine, generally, is inconsequential to the company’s bottom line; simply the cost of doing business. But the long-term damage to public health and state budgets, both in the case of Purdue and 3M, are astronomical. I wonder if it’s just going to continue this way. What changes in legislation would be required to make safety checks proactive to potential dangers, rather than reactive to damage done?

But, until then, here’s week 2 of my favourite reads from last week:

1. All the Carcinogens We Cannot See

by Siddhartha Mukherjee

Mukherjee, as usual, does a fantastic job of explaining the mysteries of cancer, and the current detective work being done to uncover their elusive ways. This article looks into an important question — what’s causing the rise of cancer rates in young people — from the perspective of carcinogens, their history, how they’re detected, how they might escape detection, and how researchers are growing their understanding of this complex issue.

It was a little unnerving to learn how little we have uncovered about carcinogens. I read this while I was in California, and every shop you enter seems to have a warning sign somewhere in it, stating that this or that substance may or may not be a carcinogen and so be careful. It’s peppered in so many places that its warning seems to have lost meaning, as if I’m supposed to hold back on my coffee or sandwich because they may have some carcinogenic elements. But, after reading this, even that seems like an under-representation.

The Ames test, which is what’s used to test whether a substance is carcinogenic or not, measures mutagenicity (the ability to cause mutations in cells). But, as this article dives into, that measurement is not comprehensive (which was known back when it was created in the 1970s as well). Carcinogens sometimes only infect in pairs: one being the “promoter”, prepping the environment (like causing inflammation, for example) so that another agent can come in and develop tumors. Reverse the order, and there are no tumors. One example of this promoter-executor (I made up the term executor) is DMBA and croton oil. DMBA, a known carcinogen, primes the cell, and when croton oil is then applied, tumors develop. By itself, croton oil doesn’t cause tumors, and passes the Ames test because its not mutagenic. Only when combined does it cause issues.

This begs the question: how many more combinations are there out there that can cause cancer. With air pollution, for example, how much higher is the risk if we looked at it through the lens of this promoter-executor model, versus just testing the effect of a single substance on cells? What if, as Mukherjee puts it, “there’s a universe of promoters” out there? (He looks into ongoing research answering this question too).

Scary, but very interesting. I wonder, with this multivariate statistical problem, can AI play a role in finding problematic promoters in this universe?

2. Private Health Care Is Here

by Christina Frangou

For years, I would constantly read about the failing healthcare system in Canada, where I live. The headlines were incessant: people getting sent back from the ER after waiting long hours, year(s)-long waits for surgeries and specialist appointments, citizens going abroad for medical procedures, shortage of prescription drugs, and so on. A few months ago I got interested in why this is happening: how did a country, known worldwide for its world-class, equitable healthcare system, let it all fall apart? I read a few stories, took a few notes, but didn’t dive too deep.

Last week I suddenly found myself interested in this topic again and this time, decided to look into it a little more. I searched for some recent articles that explained the problem, and found this one by Christina Frangou to have a good mix of history, legislation, economics, and personal stories, all wrapped to explain one result of a failing public healthcare system: people venturing to private options. Combined with a few other Maclean’s articles, I learned a bit more about the confusing half-federal, half-provincial system, the legislative loophole that allows private practices, the horrifying work environment for nurses, the shortage of nurses and family doctors, and Canada’s lack of emergency medical supplies and loss of domestic production of pharmaceuticals.

Most of the articles I’ve read do a great job of reporting problems and publicizing medical horror stories, but they generally don’t provide a list of potential solutions and how it might be implemented, except for the catch-all “more investment” (I also probably need to look more into this). I think a useful way of looking at this is to list out what health outcomes would be ideal for Canada: what features do we want (fast access to diagnostics, more family doctors, short/no wait times, etc.), and then work backwards to how to we fix, modify, create, re-create, or purge facets of the system to move towards this ideal. Essentially, an answer to the question: What does the best public healthcare system look like?

This is a really, really hard question to answer. It will involve federal and provincial politics, legislation, finance, labor markets, investment, technology, supply chain, R&D, culture and behaviour, and probably more. But I want to take a crack at it in a future essay.

3. Notes on Craft

by Greg Jackson

Greg Jackson does a great job putting into words a feeling that’s really difficult to express clearly: the foggy gap between recognizing craft and being able to create or develop your own. I find, like many others I presume, that I can recognize great writing, the way it makes me feel, but when I try to write something of my own, even when I try to mimic, it doesn’t have nearly the same effect. Understanding an element of craft in principle, even seeing it executed in practice, offers scant guidance to its artful application.

So how do you go from seeing craft to executing in practice? Jackson echoes a lot of the things I read in Anne Lamott’s incredible book, Bird by Bird: be vulnerable, watch what grabs your attention and what doesn’t, let your biases “amalgamate into your style”, don’t be afraid of failure (critique your work, constantly re-work or rewrite stuff if it’s not good enough). Great tips that they are, they’re incredibly hard to put into practice. But again, craft takes time.

A Surf Legend’s Long Ride

by William Finnegan

This is a beautiful, poignant profile of Jock Sutherland, a legendary Hawaiian surfer, by one of my favorite writers, William Finnegan. If you can only read one of the essays mentioned here, and you have no interest in surfing, I’d still recommend this. Finnegan himself is an extremely skilled surfer, so you get his characteristic eloquence in describing waves and surfing techniques, as well as his acute perceptiveness, which, when combined, paint a candid picture of Jock’s life among the waves and roofs of the North Shore and the Oahu wilderness.

At 75, he still surfs in the same spot he did since the 1950s. He also continues to work as a roofer, first starting in the 70s, and seemingly as comfortable among their steep inclines as he is in the middle of a two-storey swell. But it’s much more than surfing. The essay is a portrait of craft and obsession and love and community and skill, which results in, as Finnegan puts it, a “demonstration of basically incomprehensible mastery”. Luckily, it’s written by someone who also possesses “incomprehensible mastery”, so this might be as insightful as it gets. Bask in its beauty.

(Related story: I was sitting in the café inside the Legion of Honor, William Finnegan’s memoir, Barbarian Days: A Surfing Life on the table (I’m re-reading a favorite), when an elderly man who had been sitting a couple of tables over came by. He pointed at the book and informed me, a joyous spirit in his voice, that the upcoming issue of The New Yorker is going to have an article written by Finnegan, about the legendary surfer Jock Sutherland, who’s 75, and still surfing. I, still a little startled, excitedly told him that I had in fact seen the article on their website and had saved it to read later. A short little back-and-forth later, as he made his leave, I thanked him for coming over to tell me. I wish I could tell him that that small interaction made my day. Now, after finishing the article, I wish I could tell him how incredible it is.)

How Will Nanomachines Change the World?

by Dhruv Khullar

A really well written look into the current state of nanomachines; a good combination of historical context, descriptive behind-the-scenes reporting of scientists’ work, and commentary on the potential benefits and risks. Examples include using nanomachines to invade and kill antibiotic-resistant bacteria, harvesting rare metals from seawater, and the terrifying possibility of it starting to display emergent properties.

One particularly bonkers thing I learned was that, in 2006, a chemist named James Tour created the world’s first motor-propelled “nanocar”, roughly the width of a single strand of DNA. Just like a car has wheels connected to an axle and chassis, this nanocar had “buckyballs” — round formations of carbon — connected to a hydrogen and carbon axle and chassis. The car is “driven” by shining a UV laser on it, “causing the motor to spin, the wheels to rotate, and the vehicle to speed forward.” That’s not all. In 2017, there was an international nanocar race in the South of France, where “scientists peered at their creations using a scanning tunnelling microscope and cheered them on.” (Tour’s team won; their nanocar raced at an average speed of 95 nanometres per hour.)

I mean, this would be shocking news to me if this race happened last week, but it was 7 years ago. To make this clearly entertaining event more mainstream, my pitch is to create a spin-off of Drive to Survive, where scientists create and watch nanocars race to enter a lethal antibiotic-resistant bacteria and kill it — so, you know, literally drive to survive. Anyone have contacts in Netflix?

Mass General Cancer Center Researchers Identify Gene That Helps Cancer Cells Spread Throughout the Body

by Liz Murphy

When silenced, the Gstt1 gene that researchers found prevents cancer cells from spreading by inhibiting their ability to alter the surrounding environment to maximally benefit their growth. This prevents tumors, essentially, from being able to build a comfortable home where they can thrive. The finding opens the door to potential novel therapeutics, particularly against the dangerously venturous forms of pancreatic cancer.

research as leisure activity

by Celine Nguyen

This is a wonderful essay about trying to find joy in intellectual discovery, which can sometimes turn into a daunting task. It goes over, firstly, what research roughly is, what it means to do it for leisure (versus, say professionally or academically), and what qualities ‘research as leisure’ entail. When reading something, I often finding myself enjoying the initial discovery of something new — “woah, this is so interesting” — but later when I dive down the rabbit hole, follow the diverging threads, learn more details and try to write a coherent essay on the topic, I get bogged down, overwhelmed by just how much I don’t know or how much detail there truly is, and want to fling whatever awful draft I have into the ocean. This is a great resource and reminder — with tons of wonderful examples — on how that process, which will often be challenging, can and should nonetheless be fun.

I felt this way when I saw the structure and learned about the functioning of E.Coli in David Goodsell’s book, The Machinery of Life. Exploring, in detail, how a simple bacterium functions not only corrected some assumptions I had about cellular structures, but made me realize I had subconsciously held them in the first place, and, occasionally, never considered the specifics at all.

Take, for example, molecules moving inside a cell. I know that a cell disposes off waste; that messenger RNA goes to a ribosome to get translated into proteins; that enzymes bind to molecules to enable or disable reactions. But how, exactly, do these molecules move — how do they get where they need to go? I hadn’t really considered their inter-cellular journey. In my mind, I imagined they picked the fastest route to their destination, much like I would walking to a new coffee shop using Google Maps. But cells are a crowded place. They’re stuffed with molecules large and small: proteins, nucleic acids, amino acids, sugars, ATP, and other small molecules, all surrounded by water. Moving inside a cell is less like walking along a street, and more like bumping your way across a crowded nightclub floor. If you’re an enzyme looking for your target molecule, it’s more akin to hoping you’ll bump into it somewhere in the crowd, rather than getting to exactly where it is.

Finding out that the molecular interactions that sustain life happen by random chance, rather than intentional direction, felt a little troubling at first. I mean, chance? But again, the details reveal why this might be beneficial. Molecular interactions occur at specific orientations: the atoms or molecules connect best at some specific region, like two adjacent puzzle pieces. The crowded environment means two elements will spend more time next to each other, shuffling and bumping, which increases the likelihood of reactions. Crowding, and the barrage of bumping, also favours the assembling of molecules into larger complexes. It also means that molecular targeting (by enzymes, say) needs to be very specific — “a perfect match of shape and chemistry” — lest they react with some other molecule before running into the right one. Additionally, research by Penn State has shown that molecules move faster in crowded environments as long as the crowding is not uniformly distributed. That is, when there’s a region that’s more crowded than another nearby region, molecules are pushed towards the less crowded region faster than if the whole space was uniformly crowded or uniformly sparse. Chance, it seems, has more tricks up its sleeve than one might imagine.

Molecular transportation is a complex topic, and this method — diffusion — is just one of several ingenious ways molecules move around inside cells. There’s more to diffusion itself too, but another beautiful part of learning the details is that details tend to resemble Russian nesting dolls — there’s usually another layer to discover. This next layer, I’ll perhaps leave for a another essay.

How to (Not?) Store DNA

Another aspect of E. Coli that taught me something new is how DNA is organized inside the cell. The double helix shape of DNA is one of the most well-known biological shapes in the world. Ask a stranger what DNA stands for, and you might get a puzzled look; ask a stranger to draw DNA, and you will likely get some variation of two lines coiling around each other. I first learned about DNA in school, and its key aspects have been reinforced through nearly every biology-related article or book since. I know it consists of two complementary base pairs, adenine (A) and thymine (T), and guanine (G) and cytosine (C). I know it has a sugar-phosphate backbone. I know that it’s long — really long (over 4 million base pairs in E. Coli; approximately 3 billion in humans). And I know that DNA is primarily stored in the nucleus of cells (or in the nucleoid for prokaryotic cells like bacteria).

What I had not considered, however, was how it was stored. The nucleoid in E. Coli is a tightly-packed place, and its twisting and turning DNA, in order to fit in a space one hundredth its size, is squished in. The squishing results in something that I found as startling as I did astonishing: DNA, the most important molecule in this bacterium’s quest to reproduce, is not organized neatly, as one would imagine, but kept in a tangled, knotted mess. Looking at Goodsell’s beautiful watercolour below, it’s difficult to picture how any of the processes that involve DNA — reading the encoded information, translating a segment into RNA, splitting the strands for replication — happen at all.

But, again, life has developed some clever, if not mildly concerning, tricks. There is an attempt at organization: to facilitate the squishing, a number of curious-looking proteins are employed. There’s SMC, resembling a twisted safety-pin, that holds together loops of DNA, and HU and Fis, that act as clips to ensure two sections stay together. The result isn’t going to impress cable management enthusiasts, but it’s seemingly functional.

Even more shocking than this wacky attempt at organizing is how E. Coli resolves tangles and knots to enable those vital DNA processes. An enzyme, DNA topoisomerase, untangles DNA segments by cutting it, “allowing strands to pass by each other”, and then reconnects them after. Pause, for a second, to consider how ridiculous this is: cutting and resewing as a solution to molecular tangles. (If only such a device existed in the macro-world — would’ve been really handy when my wired headphones got tangled in my pockets all those years ago).

In the off chance that it makes an error, the cell also has an audit and repair squadron — enzymes and proteins that search and fix DNA damage — operating continuously. One of the simpler ones, the MutM protein “searches for damaged guanine bases and removes them before they can lead to a mutation.” If the damage is more extensive, the RecABC system can “repair breaks in DNA by matching the broken DNA with an intact strand”, like a quality assurance analyst, or an editor performing a spell-check. (Which leads to a question: How does it know that the “intact” one is correct?)

So far, I’ve just written briefly about two phenomena in a segment of a single E. Coli cell. A lot of cellular trickery is preserved along species, but the living world is a diverse place, and a lot isn’t. Imagine what you might learn looking closely at rose petal cells, human heart cells, falcon eye cells, jellyfish cells. How many of these stupefying, astonishing — miraculous — cellular tricks are happening in my cells right now? I wonder: do they all also cut and sew back their tangled DNA? And do their molecules also move by bumping — dancing — across their crowded inner world? Perhaps these nesting dolls are for another day.

Subscribe now

Subscribe now

I’ve always been reluctant to publish my writing. Here’s how it goes: I read something, I think of some topic to write about, either as an exercise to learn deeper or after making some connection, I begin writing a first draft, and then… that’s as far as it goes. Maybe, occasionally, there’s a second or third draft, or a back-and-forth with an editor I’ve reached out to, but only a few times have I actually published and shared it with the world (all of which have led to unexpected positive outcomes).

One might think this would encourage me to publish more, but strangely, I haven’t. I always feel that what I’ve written is not good enough, not polished enough, and so it remains in my drafts, collecting dust.

That’s why I’m making this Substack: a space to publish more early-stage, shorter-form writing — essays as I learn new things. My hope is that it might introduce some new topics or reveal new connections to readers, or see a familiar subject from a different perspective. Additionally, I hope the process will help me build up to create more polished, longer-form writing.

I’ve mostly been learning a lot of biology and biotech stuff recently, so initially that’s mostly what I’m going to write about. I have a few drafts of essays on cellular biology, science history, medicine, drug development, and scientist biographies, among other notes. I work(ed) as a software engineer so occasionally there may be some programming-based writing as well. But mostly, I’m excited to see where this leads in terms of exploring ideas and developing as a writer, and just as importantly, meeting people who share similar interests.

The writing will be free, and I aim to publish 1-2 times a week. I hope you like it!

Subscribe now