Building a Virtual Cell with AI: Key Priorities and Opportunities

Experts from Stanford University, Genentech, and the Chan-Zuckerberg Initiative are collaborating to create the virtual human cell, leveraging artificial intelligence (AI). The project aims to improve our understanding of human biology and enhance in silico experimentation, which could speed up research and support personalized medicine. By combining advances in AI and large-scale biological data, the team sees a unique opportunity to build a model that mimics the complex behaviors of human cells and tissues. This innovation holds great potential for advancing medical studies and treatments.

Digging Deeper Into the Unknown of Human Cell Modeling

Researchers are developing the world’s first virtual human cell, an AI-powered model designed to simulate the behavior of human biomolecules, cells, and tissues. This breakthrough could offer new insights into biological processes and accelerate medical research. 

“Modeling human cells can be considered the holy grail of biology,” said Emma Lundberg, associate professor of bioengineering and pathology at Stanford. She is also a senior author of a new article in Cell advocating for a global effort to create the world’s first AI virtual cell. “AI offers the ability to learn directly from data and to move beyond assumptions and hunches to discover the emergent properties of complex biological systems.”

Lundberg co-authored the article with Stanford colleagues Stephen Quake, professor of bioengineering, and Jure Leskovec, professor of computer science. Other senior authors include Theofanis Karaletsos, head of AI for science at the Chan-Zuckerberg Initiative, and Aviv Regev, executive vice president of research at Genentech. This collaboration brings together leaders in bioengineering, AI, and medical research.

The Promise and Vision of AI Virtual Cell Models

AI-powered Virtual Cells promise a breakthrough in biology and medicine. These models simulate cell behavior and molecular interactions. With advancements in AI and massive biological datasets, researchers can explore cellular functions like never before. A synthetic cell model deepens our understanding of forces that allow healthy human cells to function. It also reveals disease causes that lead to cell dysfunction or death. Furthermore, scientists can experiment in silico, using computers instead of living cells. This shift could accelerate research in human biology and speed up the search for new therapies.

The vision for AI virtual cells goes beyond a single field. For example, cancer biologists could model how mutations turn cells malignant. Microbiologists might predict how viruses affect infected cells and their hosts. Physicians could test treatments on “digital twins” of patients, leading to safer, more cost-effective personalized medicine. This idea could transform medical research by enabling simulations that offer valuable insights into human health. The ability to predict and model biological interactions would revolutionize medical practice and disease management.

However, for AI virtual cells to succeed, they must meet three key objectives. First, they must create universal representations across species and cell types. Second, they should accurately predict cellular functions, behavior, and dynamics while comprehending mechanisms. Lastly, the virtual cell should enable experiments on computers to test hypotheses and guide data collection. This would expand its capabilities faster and cheaper than today’s methods, making it a critical tool for future scientific research.

The Challenge and Path to the AI Virtual Cell

AI has introduced powerful tools that are predictive, generative, and query-able, creating new possibilities in scientific research. However, building the AI virtual cell requires handling an immense volume of biological data. The authors highlight the scale of this challenge by referencing the Short Read Archive, a DNA sequencing database maintained by the National Institutes of Health. With over 14 petabytes of data, this archive is now a thousand times larger than the dataset used to train ChatGPT.

Reaching the goal of creating an AI virtual cell will not be a simple task. It demands global collaboration on an unprecedented scale, involving experts from fields such as genetics, proteomics, and medical imaging. Close cooperation will also be needed between stakeholders in academia, industry, and non-profit organizations. The authors stress that any work toward this goal should prioritize open-access models for the scientific community to use without restriction.

“This is a mammoth project, comparable to the genome project, requiring collaboration across disciplines, industries, and nations, and we understand that fully functional models might not be available for a decade or more,” Lundberg asserted. “But, with today’s rapidly expanding AI capabilities and our massive and growing datasets, the time is ripe for science to unite and begin the work of revolutionizing the way we understand and model biology.”

AI Advances Enable Creation of Virtual Cells Simulating Molecular and Tissue Behaviors

Cells are vital for understanding health and disease, but traditional models often fail to simulate their function accurately. This recent advancement in AI and omics now presents a new opportunity to create an AI virtual cell (AIVC). This model, powered by a large neural network, can simulate the behavior of molecules, cells, and tissues in various states. Essentially it is anticipated that the development of AIVCs will transform biological research by enabling high-fidelity simulations. These models will accelerate discoveries, guide experimental studies, and provide fresh insights into cellular functions. Additionally, they will promote interdisciplinary collaborations in open science, offering exciting potential for the future of biological studies.

Author

Bernice Lottering