The Scientist Who Plans One Billion Experiments to Create a Virtual Human Cell
A scientist was born in a small town in Switzerland. Her parents were not into science, but she became fascinated with nature around her and how humans worked as humans and their biology. She really wanted to find a way to get into a lab to do some science. It was tricky for her, but eventually she talked one of her science teachers into convincing a colleague to let her into the lab. With that first science project, she went on to win the national competition and then the European Union competition. That gave her the confidence to continue with science ever since.
Why Alzheimer’s Disease Has Remained Unsolved
This scientist has been doing science for more than 20 years. Since her undergraduate days in Switzerland studying biology and neuroscience, she learned about Alzheimer’s disease. She discovered that there are big changes happening in the brain, but a lot of what is known is about late stages of the disease and how severe it is. The lecture ended with a troubling conclusion: no one really knows how it is starting. There is still no therapy. That was 17 years ago, and it stuck with her.
Why did no one understand how it is starting? Why was there no therapy? These questions drove her interest in disease biology, specifically complex diseases. Alzheimer’s is a complex disease. This does not just mean it is complicated. It means there are multiple different risk factors. Every patient has a unique combination of risk factors for the disease. This is different from an infection where you have one cause. Heart disease, many cancers, stroke, and Alzheimer’s are all complex diseases. They have a combination of genetic changes and environmental factors. Each patient is unique. The scientific community has been struggling to understand what all these different patients have in common that could be targeted to fix the disease.
Three Breakthroughs That Change Everything
Now, this scientist sees an opportunity for a different kind of attack on these diseases. Three things have come together just in the last one or two years that make it possible to understand such complex problems. These three areas are measuring, changing, and understanding.
Measuring means single-cell sequencing. This technology allows researchers to look at one cell at a time and take a snapshot of key dynamic processes in the cell, which is the RNA expression. RNA is like the language of the cell. This takes a snapshot one cell at a time of what is going on inside it.
Changing means having the ability to change something very precise. This involves changing one gene at a time to stop it from making RNA or changing it to upregulate the RNA. The scientist has been working on CRISPR technology for 15 years. The field has made many advancements, and now scientists can do this across all the genes in the genome. They can make these changes in a targeted way, which has only become possible very recently.
Understanding is where AI comes in. Just as AI has cracked understanding human language, the scientist sees the possibility that AI can be used to understand the language of our own cells, RNA. This is the core principle. To do this, you need to be able to measure it and change it in a targeted way.
Why RNA Language Is Different from Human Language
Six years ago, people were not sure that large language models could scale to build a conception of the world and approximate intelligence. But the key insight was that a model can learn so much just from human language. Similarly, that concept can be applied to RNA, which is the language of the cell, especially the dynamic language because it is changing all the time. It reflects what is happening to the cell and also reflects the cell’s genetics.
Human language is generated by humans, so humans understand it. RNA language or biological language has evolved. It was not generated by humans. It is basically impenetrable for humans. Humans can predict the left side: “to be or not to be” from Shakespeare and complete it. But no human could complete the RNA sequence on the right side. AI does not care about this difference. AI can learn biological language even though humans cannot understand it.
The Plan for One Billion Experiments
Large language models are very data hungry. They have been trained on all human languages generated over generations and civilizations. In biology, there is nothing similar to that. Scientists need precise measurements one cell at a time. They also need to know what actually happened to that cell to build a predictive, dynamic model that can predict how a cell will change when something happens to it. This data set needs to be generated.
The scientist’s plan is to do at least one billion experiments over the next four years. These are all physical experiments, not software simulations. She is an experimental biologist. The team uses tricks to make this scalable, using bar-coding technologies to run experiments in bigger pools and then back out what happened to the cells. They have already done about 60 million experiments so far, so they feel confident they can keep going.
How the Virtual Cell Will Work
The whole point of generating this model is ultimately for human health. For a disease state, such as a certain cell in Alzheimer’s disease like an immune cell in the brain called microglia, scientists can measure what that looks like not just for one patient but across many patients. This data is already out there, so they do not even have to generate it. They can see all the diseased cells and all the healthy cells across people. Then they can ask the model: the model knows how to change cells, so what intervention, what genetic change, what chemical change is needed to convert all the diseased cells across all patients with the same disease back to healthy cells?
If someone truly understands the language of DNA, the model can predict something medicine has never known before. The answer might be a complex series of interventions needed for that cell. It is not as simple as giving an aspirin. It could be a complex combination of things, or it could be a question of picking the correct one out of 20,000 possibilities that could be up or down regulated, totaling 40,000 possibilities.
Normally, target identification in biomedicine works through a guess and check approach. A researcher has a hypothesis about one gene and spends a few years checking whether it is the right one. With 40,000 things to pick from, even picking a single one takes forever. That is why these diseases have not been cured yet.
The Universal Virtual Cell Vision
The whole point is that it is a universal virtual cell. It needs to learn how to generalize to a new kind of cell or a new state of a cell, a new disease, without having seen training data for that new cell type. This is a very challenging task. The team is thinking hard about how to do these experiments.
The vision is real. The team has already built their first model that came out eight months ago. It is not very good. It is state-of-the-art and the best model at the time it was published, but it has a long way to go to be at the accuracy needed to be truly useful. The interface using that model allows users to say they have a cell they are starting with and want to change something about the cell. The model then spits out different changes that can be made to the cell that are most likely to shift it the way the user wants.
Making the Tool Available to Everyone
The scientist is not holding on to this tool or licensing it to companies. She is making it generally available. There are a few ways people can interact with it and follow along. The tool will be released later this year for people to try. There will be caveats that it is not very accurate, maybe 20 percent accurate, but the team will iterate over the next four years. They are also hosting a Virtual Cell Challenge every year for the whole community. The first one had 1,000 teams participating, which helps move the whole field forward.
Addressing Safety Concerns
Some people might wonder if making this tool available is dangerous. Could someone use it for harmful purposes? The scientist says the tool is really just for human cells. In theory, someone could build this kind of tool for a virus, but that would be a bad idea because then it could be used to create something dangerous. However, this tool only allows shifting human cells into a different state, which would be pretty difficult to abuse. If a nasty virus did come along, the model would actually help defend. It would show how the virus is targeting a gene in a cell and what happens to the cell when that gets targeted, so it would help researchers defend against it.
The Arc Institute and Its Growth
Arc Institute was started in 2022, so it has only been four years. It has grown a lot. The before and after picture shows just one year of growth. The institute now has over 300 people. The goal was to bring people together from different disciplines and have AI and biology under one roof. They started just around the right time when they could see what machine learning was going to mean for biology.
A Message for Families Affected by These Diseases
For someone who has a family member with Alzheimer’s or heart disease, the scientist says medicine is going to transform for these kinds of diseases. Maybe not in three months, so patience is needed. But within four or five years, there will be models that are accurate enough to be useful. It is a totally different way of doing biology, not one hypothesis at a time. A field like Alzheimer’s can get bogged down by focusing on one dominant hypothesis that might be wrong. With these models, researchers can take a comprehensive data-driven look at all the things that could be targeted with a drug, what will happen with all of them, and which one will be the most effective. It is a totally different way of tackling the problem that is incredibly exciting.
Conclusion
This scientist’s vision represents a revolutionary shift in how we approach complex diseases. By combining single-cell sequencing, CRISPR technology, and artificial intelligence, she aims to create a universal virtual cell that can predict how cells respond to changes. The plan to run one billion experiments is ambitious but achievable with the right technology and scaling methods. Making this tool freely available to the global scientific community accelerates research worldwide. While safety concerns exist, the focus on human cells and the defensive potential against viruses makes this a net positive for humanity. For millions of families affected by Alzheimer’s, heart disease, and other complex conditions, this research offers genuine hope that effective treatments may finally be within reach. The transformation of medicine is coming, and it starts with understanding the language of our own cells.