Recently, a research project jointly conducted by Stanford University, Genentech and the Chan Zuckerberg Foundation published a forward-looking opinion article in the journal Cell. The focus of the study is to develop the world’s first virtual human cell, aiming to open up new paths for understanding and simulating the functions and behaviors of cells through the power of artificial intelligence.
The Functions of Artificial Intelligence Virtual Cells are Diverse
Cells, as the basic building blocks of life, play a vital role in understanding health and disease. However, traditional models have limitations in modeling and simulating cell functions. In response to this dilemma, researchers have proposed a new idea that combines AI and omics technology. This study proposes an innovative solution- artificial intelligence virtual cells (AIVC), which combines artificial intelligence and omics technology to provide unprecedented possibilities for simulating cell functions.
The core of AIVC lies in its multi-scale and multi-modal model construction, which can not only simulate the complex interactions inside cells more accurately, but also has the potential to accelerate the progress of scientific research. Through AIVC, researchers can more effectively guide experimental research and predict cell behavior, thereby promoting breakthrough progress in the field of biology. This technology also promotes interdisciplinary collaboration and provides new opportunities for the integration of fields such as biology, computer science and medicine.
The writing team of this study brings together many top scientists, including Professor Stephen Quake and Professor Jure Leskovec of Stanford University, who serve as the scientific director of the Chan Zuckerberg Initiative and professor of computer science in the School of Engineering, respectively. Theofanis Karaletsos, director of scientific artificial intelligence at the Chan Zuckerberg Foundation, and Aviv Regev, executive vice president of research at Genentech, also participated in the writing of the article as senior authors.
AIVC is versatile and powerful. It can not only simulate the response of cells under specific conditions, but also predict the changes of cells in disease states. The development of this technology provides researchers with a powerful tool to enable them to gain a deeper understanding of the mysteries of life and provide new ideas for the prevention and treatment of diseases.
This synthetic cell model can not only allow people to have a deeper understanding of the complex interaction of chemical, electrical, mechanical and other forces and processes that make healthy human cells work, but also reveal the root causes of diseases that cause cell dysfunction or death.
Perhaps more interestingly, AI virtual cells will also allow scientists to conduct in silico experiments on computers rather than on living cells and organisms, rather than in vivo. This capability will expand our understanding of human biology and accelerate the search for new treatments, drugs, and even cures for diseases.
Visions of the Digital Biology Era
Cancer biologists might simulate how certain mutations turn healthy cells malignant. Microbiologists might one day predict the effects of viruses on infected cells and even host organisms. Doctors might one day test treatments on a patient’s “digital twin,” accelerating the long-promised era of faster, more affordable, and safer personalized medicine.
For AI virtual cells to be successful, though, they need to achieve three goals. First, it needs to give researchers the ability to create universal representations across species and cell types. It must also accurately predict cellular functions, behaviors, and dynamics, and understand cellular mechanisms. Last but not least, AI virtual cells will allow experiments to be conducted in silico to test hypotheses and guide data collection, expanding the capabilities of virtual cells at a much lower rate and cost than is currently possible.
The Challenge of Massive Biological Data
AI has ushered in an era of predictable, generative, and queryable tools, but it is undeniable that creating virtual cells requires a large amount of raw biological data. In contrast, the DNA sequencing database compiled by the National Institutes of Health, called the Short Read Archive, currently contains more than 14PB of data, a thousand times larger than the dataset used to train ChatGPT.
Achieving AI virtual cells will not be easy. It will require a collaborative, open science collaboration on a global scale, involving fields ranging from genetics to proteomics to medical imaging, and close cooperation between global stakeholders such as academia, industry, and nonprofit organizations. At the same time, any work done for AI virtual cells must assume that the resulting models will be open to the entire scientific community without restriction.
“This is a massive project, comparable to the genome project, requiring collaboration across disciplines, industries, and countries, and we know that fully functional models may not be available for a decade or more,” Lundberg asserted. “But with the rapid advancement of today’s AI capabilities, and our large and growing datasets, the time is ripe for the scientific community to come together and begin to revolutionize the way we understand and simulate biology.”
