Why AI Companies Are Racing to Build a Virtual Human Cell

The human cell is a Rube Goldberg machine like no other, full of biological chain reactions that make the difference between life and death. Understanding these delicate relationships and why they occur in disease is one of the major challenges of biology. A single mistake in a gene can cause the protein it makes to take the wrong shape. A deformed protein will not be able to do its job. And due to the lack of this protein, the body – you – can begin to fall apart.

However, cells are so complex that it is difficult to understand how the failure of one protein spreads throughout the system. Graham Johnson, a computational biologist and scientific illustrator at the Allen Institute for Cell Science, remembers fantasizing at the dinner table more than 15 years ago about a computer model of a cell so detailed and complete that scientists could watch such processes take place. At that point, “everyone just chuckled,” he says. “It was too surreal.”

But now some researchers are using AI to take new steps toward creating a “virtual cell.” Google's DeepMind is I'm working on a project like thisand the Chan Zuckerberg Initiative (CZI) has made virtual cells a major focus of its Biohub research network, says Theo Karaletsos, senior director of artificial intelligence at CZI. There are even new prizecreated by Arc Institute for virtual cell style models. The goal of all these efforts is to predict how both healthy and diseased cells work in such detail that drug development can be accelerated and scientific discovery can be accelerated. Some believe virtual cells could even simplify basic research by moving biologists from the lab bench to the keyboard.

What is a virtual cell?

The exact definition of a virtual cell depends on who you're talking to. Some scientists, like Johnson, hope the virtual cage will include a visual representation that you can click on and explore. Others think of it primarily as a set of computer programs that can answer questions and make predictions about what might happen. But this concept is not a new idea. For decades, biologists have created mathematical models of cellular processes. To create them, researchers rely on data from experiments with real cells and create equations that describe what happens.

There is now more data about the human cell than ever before, thanks in part to technology that allows scientists to spy on the activities of individual cells. But figuring out the equations for each process and putting them all together is a monumental task. “The old way of doing this—by hand—had, I would say, only very limited success,” says Stephen Quake, a professor at Stanford University and former chief scientific officer at CZI. Last year he and other researchers published an article outlining the concept of a different approach that feeds cellular data directly to specialized AI. “You build models that learn directly from the data, rather than trying to write equations,” he says.

Read more: Why are baby boxes suddenly everywhere?

Quake and his colleagues had some interesting early results. To train the AI, they used data from cells from 12 different species. According to Quake, the AI ​​was then able to make accurate predictions about the cells of species it had never seen before. He was also able to infer relationships between different types of cells of the same species, despite the fact that he was not given any information about these relationships. “Personally, I was very excited about this approach,” says Quake.

Another group of researchers, including from Google DeepMind, exploring the use of AI to create virtual cells. They trained the AI ​​to work with large datasets of cells, allowing users to ask questions like: “How will this cell respond to this drug?” and then get answers about what parts of the cell might be affected.

These are just some of the approaches scientists are using to create virtual cells. It is likely that there will eventually be many different types of virtual cells designed for use by different researchers. For example, the virtual cell used by a cancer biologist may be different from the one used by a cell biologist seeking to answer questions about how a given structure evolved. And it is quite possible that they will use both traditional modeling approaches and artificial intelligence.

What virtual cells can allow us to do

Virtual cells could speed up and simplify the discovery of new drugs. They can also provide insight into how cancer cells evade the immune system or how an individual patient may respond to a given therapy. They may even help basic scientists come up with hypotheses about how cells work, which will help them conduct experiments with real cells. “The overall goal here,” says Quake, “is to try to transform cell biology from being a field that is 90% experimental and 10% computational into the opposite direction.”

Some scientists wonder how useful AI predictions will be if AI can't explain them. “AI models tend to be a black box,” says Eric Armingol, a systems biologist and fellow at the Wellcome Sanger Institute in the UK. In other words, they give you an answer, but they can't tell you why they gave you that answer.

Read more: 11 things therapists wish every child knew

“Personally, I got into this field because I wanted to model the entire human body and how cells connect to each other and interact. So it's a dream,” he says. Black box responses may be useful for guiding drug development, but they may not be as useful for basic scientists—at least not the way many AIs are currently designed. (CZI's Karaletsos says some of their AIs are designed to provide explanations for their reasoning. “We want to understand, not just predict,” he says.)

Johnson, author paper in 2023 about the importance of creating virtual cells, hopes that whatever scientists eventually create can be visualized. His ethos is “a visual, interactive, intuitive version of something complex,” he says. “I think AI is absolutely necessary to enable all of this. I'm just not interested in black box predictions as the main outcome.”