New AI Accurately Predicts Gene Activity in Human Cells

Using advanced artificial intelligence, researchers at Columbia University Vagelos College of Physicians and Surgeons have developed a system that can predict the activity of genes within any human cell. Described in the journal Nature, this groundbreaking technology offers unprecedented insight into the inner workings of cells and has the potential to transform fields like cancer research and genetic disease study.

How the AI Works

The system predicts which genes are active in a given cell by analyzing gene expression data, which reveals a cell's identity and function. Traditional biological research methods excel at uncovering how cells function and respond to disturbances. However, they fall short in predicting how cells operate or how they might respond to changes, such as a cancer-causing mutation.

The approach resembles the way foundation AI models like ChatGPT learn language rules. Here’s how it was developed:

• Researchers trained the AI on data from over 1.3 million human cells sourced from normal tissues.

• The inputs included two types of biological data: the complete DNA sequences of cells (genome sequences) and information about which parts of the genome are "open" and active (accessible, expressed regions). These active regions are where genes are turned on to produce proteins or other molecules, determining how the cell functions.

• The AI learned the underlying "rules" or patterns (the "grammar") of how genes are activated in cells and used this knowledge to accurately predict gene activity in cell types it had never encountered before.

• The model's predictions closely matched experimental data, demonstrating its reliability and potential to guide further discoveries.

“Predictive generalizable computational models allow us to uncover biological processes in a fast and accurate way. These methods can effectively conduct large-scale computational experiments, boosting and guiding traditional experimental approaches,” says Raul Rabadan, professor of systems biology and senior author of the new paper.

Applications in Cancer and Beyond

The researchers tested their AI system on an inherited form of pediatric leukemia to uncover previously unknown mechanisms.

The AI predicted that specific mutations disrupted interactions between transcription factors crucial for determining the fate of leukemic cells.

Laboratory experiments confirmed the AI’s findings, shedding light on the biology of the disease and identifying potential new targets for treatment.

Additionally, the system could help researchers explore the genome’s “dark matter,” vast regions that don’t encode proteins but are linked to diseases like cancer. “The vast majority of mutations found in cancer patients are in so-called dark regions of the genome," says Rabadan. These regions remain largely uncharted, but the AI could help decode their role in disease development.

Transforming Biology Into a Predictive Science

This new AI system is part of a broader trend in biology, where vast datasets and powerful machine learning models are enabling predictive research. As senior author Raul Rabadan explains: “Having the ability to accurately predict a cell's activities would transform our understanding of fundamental biological processes. It’s really a new era in biology that is extremely exciting; transforming biology into a predictive science.”

Rabadan’s team is already applying their model to study various cancers, including brain and blood cancers, to uncover how cells change during disease progression. The potential applications extend far beyond cancer, opening doors to understanding genetic disorders, improving drug discovery, and identifying therapeutic targets.

What This Means

This breakthrough highlights the growing synergy between AI and biology, promising to make the latter a more predictive science. By uncovering patterns in cellular behavior, AI can accelerate experimental research, reveal hidden disease mechanisms, and pinpoint new treatment strategies.

As researchers go deeper into areas like the genome's "dark matter," this AI-driven approach could redefine our understanding of diseases and their underlying biology. It’s a pivotal step toward using predictive tools to combat cancer, genetic disorders, and other complex medical challenges.

Editor’s Note: This article was created by Alicia Shapiro, CMO of AiNews.com, with writing, image, and idea-generation support from ChatGPT, an AI assistant. However, the final perspective and editorial choices are solely Alicia Shapiro’s. Special thanks to ChatGPT for assistance with research and editorial support in crafting this article.