Israeli researchers develop AI tool to discover how cells react to drug treatment
Tel Aviv University team’s scNET algorithm combines data on cells with protein ‘social network’ to cut ‘high noise’ and shed light on immune cell behavior
Reporter at The Times of Israel
In a first, Tel Aviv University researchers say they have developed an innovative AI tool that integrates biological data to provide insights into how cells behave and react to different drug treatments. The breakthrough method could improve treatment strategies for diseases like cancer, they say.
The system, called scNET (pronounced sek-net), short for single-cell network, “uses AI in a slightly different way,” said Prof. Asaf Madi from the Faculty of Medicine, who worked with Prof. Roded Sharan, head of the School of Computer Science and AI, and PhD student Ron Sheinin.
Speaking to The Times of Israel by phone, Madi said the researchers developed a “computational platform to make a smart integration of this data, and the results were really, really astonishing.”
The findings appeared in the peer-reviewed journal Nature Methods.
Single cell information with a lot of ‘noise’
Madi explained that in the past decade, researchers have been able to collect data to examine single cells from different tissues, created through the process of gene expression.
Researchers were able to investigate different cell populations inside a tumor, for example, and then use these discoveries to learn about the potential functional behavior of each cell type.
“This was the revolution of research in our field in the last decade or so,” said Madi, founder of Systems Immunology Lab at the university.
The scientists could begin to search for the deep specific mechanisms of how cells behave. For example, what exactly do immune cells, known as T cells, a type of white blood cell, do to help the immune system fight germs? What is the complex role of B cells, a different kind of white blood cell that makes antibodies? And what triggers the behavior of macrophages, a third type of white blood cell, to surround and kill microorganisms?
Researchers were armed with complicated questions, but they couldn’t get the answers because, Madi said, they had no way to measure the activity in these cells due to “high noise,” the background activity that makes it difficult to zero in on specific behaviors.
Noise in genetic data is the extra information that makes it difficult for scientists to discern what is really happening inside cells.
“The early technology is great,” Madi explained, “but it isn’t perfect, and some of the information is not as accurate as you would want it to be.”
A protein ‘social network’
At the same time, in another field of study, researchers have been working to develop a protein interaction database that “reflects the potential interaction between all the proteins that we have in our bodies,” said Madi.
This research has generated “a huge map” resembling a social network, he said.
Madi and his team wanted to take the information from the data on gene expression and combine it with the data from the protein map. They believed that integrating the two domains would improve accuracy in immunology research.
The work took about two years.
“This is something that is not easy to do,” Madi said. “This was the problem that we sought to solve.”
What does scNET do?
By developing scNET, the researchers were able to get a more comprehensive picture of behavior inside the cell.
“The scNET maps out interactions between genes, showing how they influence each other,” said PhD student Sheinin. By incorporating that map with the single-cell sequencing data, scNET enabled more precise identification of cell populations and their behaviors under different conditions.
The research team then went back to a previous experiment with a cancer treatment carried out on mice in the lab, but this time using scNET.
In the first try, they had seen a delay in tumor growth and the prolonged survival of the mice, “but we could not explain why that was happening,” Madi said. “But when we used scNET, it revealed the effects of the treatment on the T cells that help the immune system fight the disease.”
The scNET showed that “the T cells were activating programs associated with fighting the tumor,” he explained. “And this was not possible to discover due to the high noise in the original data. However, it was something that scNET was able to uncover.”
Madi said the field is “moving very fast,” and there are rapid advances. He hopes that scNET can be “taken forward” and used in other areas to improve the understanding of disease and thereby accelerate treatment development.
“We really hope that other researchers will use it to revisit some of their older research, where they didn’t get such promising results from the data,” Madi said.
“We developed this method, we showed that it’s working. We really hope that others will pick up what we’ve generated and use it for their own research.”
The Times of Israel Community.