TAU develops AI tool to decode cell responses to cancer treatments

A new artificial intelligence-based method developed at Tel Aviv University could significantly enhance our understanding of how cells respond to drug treatments - especially within complex environments like cancerous tumors.

The innovative system, called scNET, merges two previously separate streams of biological data: gene activity at the single-cell level and known interactions between genes. According to its developers, this dual-layered approach allows researchers to identify subtle but critical changes in how cells behave, particularly in response to treatments such as immunotherapy or chemotherapy.

The peer-reviewed study, published in Nature Methods, was led by PhD student Ron Sheinin, under the supervision of Prof. Asaf Madi of TAU's Faculty of Medicine and Prof. Roded Sharan, head of the university's School of Computer Science and Artificial Intelligence.

"Today's technologies give us unprecedented resolution into what individual cells are doing," Madi said. "But the data is often noisy, which makes it hard to draw clear conclusions - especially about rare but important cell populations like tumor-fighting immune cells. That's where scNET comes in."

How does it work?

scNET uses AI to overlay raw gene expression data with a kind of "biological social network" - a map of how genes are known to interact and influence one another. This network-based approach, Sheinin explained, "lets us identify gene activity patterns that were previously hidden in the noise. We can now see how immune cells like T cells ramp up their activity in response to a treatment - something that was nearly impossible to detect before."

The researchers specifically applied the tool to T cells, a key component of the immune system known for their ability to attack cancer cells. In treated tumor environments, scNET revealed previously undetectable increases in T cell cytotoxicity - their capacity to destroy cancerous cells.

“This is a powerful demonstration of how artificial intelligence can help decipher biological and medical data,” said Prof. Sharan. “We are giving researchers computational tools that allow them to see the bigger picture - and find answers that might otherwise be missed.”

Beyond cancer research, the team believes scNET could have broad applications in the development of new treatments, better understanding of immune function, and personalized medicine.

“This is just the beginning,” Sheinin added. “Our framework can be used to investigate many types of diseases and potentially guide clinical decisions based on how individual cells respond to therapy.”