The Digital Worm: Can We Recreate Life One Neuron At A Time?
Gusts of the Santa Ana winds were already howling when I launched my first worm simulation. I’m no hacker, but the process was surprisingly straightforward: pull up a Terminal shell, paste a handful of lines of code copied from GitHub, and watch characters stream down the screen just like they do in every 90s hacker movie. I was scanning the rolling lines of code for terms I recognized—neuron, synapse—when my friend pulled up to take me out for dinner. “Give me one second!” I called from my home office. “I’m just running a worm on my computer.”
The energy at the local Korean spot felt almost frantic, matching the chaotic weather outside. The wind bent palm trees nearly double and sent empty shopping carts careening across the asphalt parking lot. The whole atmosphere felt amplified, uncanny, like listening to a podcast played at twice its normal speed. “You’re doing what, a cybercrime?” my friend asked over the noise of the crowded restaurant. I shouted back to clarify: no, this wasn’t a malicious worm like Stuxnet. This was a worm more along the lines of what a biologist would study—a living creature, rendered entirely in code.
By the time I got back home, the sky had gone dark, and the first wildfire embers had already touched down in Altadena. Waiting for me on my laptop, tucked inside a 3D voxel grid, was my worm. Pointed at both ends, it hovered in a cloud of tiny particles, unnervingly straight and completely still. Obviously, it wasn’t alive. Even so, it looked far more inert than I expected. “Bravo,” Stephen Larson told me when I called him later that night. “You just got to the ‘hello world’ stage of the simulation.”
Larson is a co-founder of OpenWorm, an open-source software project that has worked since 2011 to build a complete computer simulation of Caenorhabditis elegans, a microscopic roundworm. His team’s end goal is nothing short of a full digital twin of the real worm, accurate all the way down to the molecular level. If OpenWorm pulls this off, it will be the first fully simulated animal on Earth—and a culmination of everything we know not just about C. elegans, one of the most intensively studied organisms in all of biology, but also how brains interact with their environments to produce behavior. The project calls this the “holy grail” of systems biology.
Unfortunately, they haven’t yet cracked the problem. The simulation I ran on my laptop pulls data collected from experiments on living worms and translates it into a computational framework called c302, which then controls the simulated muscle system of a C. elegans in a simulated fluid environment. All together, it models how the worm crawls across a flat layer of gel. Generating just five seconds of this movement takes roughly 10 hours of computing time.
A lot can change in 10 hours. A single ember can catch a ride on the wind, blow down from the foothills, and ignite a whole sleeping city. That night, following Larson’s advice, I adjusted the simulation’s time parameters, moving past the basic “hello world” stage and deeper into the uncanny valley of simulated life. The next morning, I woke up to an eerie, burnt-orange haze hanging over the city. Bleary-eyed, I opened my laptop, and two revelations made my heart skip a beat: Los Angeles was ablaze. And my worm had moved.
By this point, you’re probably asking a very logical question. My friend asked the same thing between bites of banchan back at the restaurant. It’s the question everyone asks: Why, in a world teetering on so many crises, with so many urgent problems begging for solutions, would anyone spend 13 years trying to code a microscopic worm into digital existence?
As a starting point for an answer, we can turn to one of physicist Richard Feynman’s most famous maxims: “What I cannot create, I do not understand.” For most of its modern history, biology has been a reductionist science, built on the idea that the best way to unpack the overwhelming complexity of living things is to break them down into their smallest components: organs, cells, proteins, molecules. But life isn’t a simple clockwork mechanism. It’s a dynamic system, where surprising, emergent properties arise from the interactions between all those tiny parts. To truly understand life, you can’t just take it apart. You have to be able to put it back together again.
C. elegans is a tiny roundworm, barely longer than a human hair is wide, with fewer than 1,000 cells in its entire body. Only 302 of those cells are neurons—about as small as a functional brain can get. “I remember when my first child was born, how proud I was when they finally learned to count to 302,” jokes Netta Cohen, a computational neuroscientist who runs a worm research lab at the University of Leeds. But Cohen is quick to point out that there’s no shame in being small: C. elegans does a remarkable lot with so little. Unlike harmful parasitic worm species, it doesn’t rely on larger host organisms to survive. It’s what biologists call a free-living organism. “It can reproduce, it can eat, it can forage for food, it can escape danger,” Cohen says. “It’s born, it grows, it ages, it dies—all in the space of one millimeter.”
Researchers who study C. elegans like Cohen are quick to note that no fewer than four Nobel Prizes have been awarded for work on this tiny organism. It was the first animal to have both its full genome sequenced and its entire neural wiring map mapped out. But a wiring schematic is not the same as an instruction manual for how the system works. “We know the connections between neurons; we don’t know how the whole system behaves dynamically,” Cohen says. “That makes it the perfect problem for physicists, computer scientists, and mathematicians to tackle.”
They’ve been trying for decades. The first person to attempt a simulation of C. elegans was Sydney Brenner, the researcher who lifted the humble compost-dwelling worm to scientific stardom with his landmark 1986 paper The Structure of the Nervous System of the Nematode Caenorhabditis elegans, reverently known among worm biologists as “The Mind of a Worm.” Working out of a lab in Cambridge, UK, Brenner’s team spent 13 years painstakingly slicing worm specimens and photographing them under an electron microscope, using a first-generation minicomputer—the kind programmed with punched paper tape—to piece their data together into a rough map of the worm’s nervous system.
Every 10 to 20 years since then, computer scientists have tried to build on Brenner’s work, but biology has a way of humbling even the most ambitious computational researchers. In 2003, computer scientist David Harel called simulating a full C. elegans a “grand challenge” for biology, arguing that the field was long overdue for a major shift “from analysis to synthesis.” But while Harel was correct about the need for that shift, he never managed to model more than the worm’s vulva—yes, that’s a true story.
Cohen herself has spent more than 20 years publishing groundbreaking computational models that accurately replicate the sinusoidal crawling motion C. elegans uses to move forward through fluids of different viscosities. But how the worm moves backward is a completely separate, unsolved problem—don’t even get started on how it moves up or down, or why it moves the way it does at all. Almost all existing data on C. elegans behavior comes from specimens studied on flat agar plates in labs; for all we know, they behave completely differently in their natural wild habitats. “Why not?” Cohen says with a laugh. “It’s biology.”
When OpenWorm announced its project in 2011, Stephen Larson, an engineer who had become a devout believer in open-source collaboration, thought that if he could bring together a team of dedicated computational researchers to tackle this biological problem, they could make real progress toward a full simulation. Thirteen years later, Larson is more reflective about the project’s pace. “Maybe this project is a cathedral,” he told me. “If I don’t live to finish it, at least other people can see what we’ve started and keep building on it.”
That perspective could come from burnout: leading an underfunded open-source project, even for a few years, can drain even the most idealistic of leaders. It could also be a testament to the deceptive complexity of C. elegans’ tiny brain, which still resists being fully captured by code. Or it could just be a matter of bad timing.
OpenWorm doesn’t conduct its own original lab experiments. Instead, the project’s network of volunteers pull existing data from decades of published C. elegans research, integrating whatever available data they can find into their simulation. That means the project is dependent on work from labs like Cohen’s, which have been slow to generate the specific types of high-quality data that computational models really need. But over the last decade, experimental biologists have upgraded their microscopes and refined genetic imaging techniques, producing more and better recordings of the worm’s brain in action. At the same time, machine learning tools have advanced dramatically to make sense of all that new data, and computing power has skyrocketed. This confluence of progress makes Larson optimistic. “When you’re living through a period of almost exponential technological growth, something that sounds crazy might actually be possible,” he says.
I asked Cohen, who serves on OpenWorm’s scientific advisory board, if she thinks it’s actually possible. “Well, let’s start from the assumption that it is,” she said. “Then we can ask what we need to do to get there.” Cohen is one of 37 co-authors of a recent position paper outlining a bold new plan: use genetic imaging technology to activate each neuron in the worm’s nervous system one at a time, measuring how that activation changes the activity of the other 301 neurons. If this process is repeated hundreds of thousands of times in parallel experiments, it will collect enough data to finally give computational researchers what they need to fully reverse engineer the entire worm.
It’s an ambitious proposal, one that will require an unprecedented level of collaboration between 20 different worm research labs around the world. Gal Haspel, a computational neuroscientist at the New Jersey Institute of Technology and the lead author of the reverse engineering paper, estimates that pulling it off could take up to 10 years, cost tens of millions of dollars, and require somewhere between 100,000 and 200,000 real C. elegans specimens. In the process, it will generate more data about C. elegans than has been collected in the entire history of research on the organism. And what will the team have to show for all that work in the end? “All these people, all these computers,” Haspel says. “And we’ll end up with something that one tiny animal can do naturally, right now.”
Haspel says that with a dry sense of humor. He compares the project to NASA’s Apollo moonshot: it’s the kind of big, ambitious project that pushes technology forward, forces engineers to build better tools and scientists to work together across disciplines. Haspel believes the worm simulation is an opportunity to build a new kind of science, one driven by automation, big data, and machine learning. And even though the end product is just a worm—an expensive, inefficient one, basically the world’s most sophisticated Tamagotchi—it can be a stepping stone toward understanding more complex nervous systems, and eventually, the human mind.
Last summer, a crypto developer posted an animated GIF on X showing a virtual C. elegans bouncing around an on-screen window. The animation was built with the same code I ran on my laptop, which is available for free on OpenWorm’s GitHub page. “If the worm matrix runs on my M1 Mac,” he wrote, “what are the odds we’re actually in base reality?” He meant that maybe we’re all simulated worms too, running on a cosmic MacBook sitting on some higher-dimension workdesk. The post went viral, and Elon Musk, predictably, liked it.
When I mentioned the “worm matrix” idea to OpenWorm’s project director Padraig Gleeson, a computational neuroscientist at University College London, he visibly winced. “Some people get into this project because they want to have philosophical discussions about that kind of thing, and that’s fine,” he said. “My priority is actually advancing the biology, first and foremost.”
Gleeson is the practical engineer to Larson’s visionary founder: he’s less interested in building a perfect “über-worm” than he is in building a flexible platform that combines smaller, more focused models of C. elegans’ biological systems. Computational modeling is already standard practice in biology; it’s a low-cost way to encode and test hypotheses as “thought experiments” before researchers break out agar plates and worm food for lab work. Usually, models only focus on one small part of an organism—for example, the handful of neurons that control forward movement. When it comes to modeling, “we don’t need the map to be as good as the territory itself. That defeats the whole purpose,” explains Eduardo Izquierdo, a computational neuroscientist at the Rose-Hulman Institute of Technology who studies worm modeling. “We need something that helps us work through our questions about how life works.”
No one mistakes a standard biological model for the real living thing. But a full, accurate simulation opens up a whole different set of questions. To borrow Izquierdo’s framing, it is a map that is just as good as the territory it represents—and that invites new speculation about the nature of that territory, and of life itself. If a model helps scientists answer questions, a simulation raises new ones. For example: what separates a virtual worm from its living counterpart, if the two are identical all the way down to the molecular level?
In Larson’s view, a fully accurate worm simulation will expand how we think about life: instead of invalidating our current understanding, it will broaden it. “If we define aliveness as something that can only exist in systems of physical molecules that have mass and exist physically on Earth, then something in a computer without physical molecules can’t be alive,” he says. “But if we expand our definition of aliveness to center on information instead, then there might be a version of aliveness that applies to a simulated animal. And at that point, does it really matter?”
I think it does matter. Life is made of information, but it’s more than that too—it’s something we feel most acutely when it’s gone. Looking at this through that lens, I wonder if Feynman’s famous dictum needs a small tweak. It’s not exactly that creating something gives you understanding. It’s that only by trying to recreate life can we come to understand how irreplaceable it is.
Of course, I say this because I’m surrounded by destruction right now. The air is toxic, and flakes of white ash have drifted into every crack and crevice of my house. Sitting just outside the evacuation zone, close enough to smell the smoke, I distract myself by running more and more proto-simulations of C. elegans. Watching them, I can’t help but marvel at how easy it is to destroy life, and how incredibly hard it is to create it. It only takes one spark to burn down centuries of growth in a single night. But to get a virtual worm to crawl just one sluggish inch forward? That’s taken decades of work, and it may never even be finished.
Let us know what you think about this article. Submit a letter to the editor at [email protected].
