Defense backs pioneering research into machine learning for biological networks

Written by
Scott Lyon
March 20, 2024

The Defense Department has announced support for a project that will use machine learning techniques developed at Princeton and collaborating institutions to understand causal relationships in large, extremely complex systems, from the networks of living cells that form tissues and organs to the interdependent behaviors that lead to global pandemics.

Mengdi Wang, associate professor of electrical and computer engineering, will lead the AI and reinforcement learning thrust of the collaborative project, which won a $7.5 million grant from the Multidisciplinary University Research Initiative (MURI). The MURI program focuses on complex problems that cut across disciplinary boundaries and supports “teams whose members have diverse sets of expertise as well as creative scientific approaches to tackling problems,” Bindu Nair, a representative of the program, said in a statement released by the Department.

The team features leading experts in biological engineering, economics, statistics, probability, epidemiology and machine learning, with researchers from the Massachusetts Institute of Technology, Harvard University, the Broad Institute of MIT and Harvard, and Stanford University. Caroline Uhler, director of the Broad Institute’s Eric and Wendy Schmidt Center, will head the project. Wang, Uhler and Feng Zhang, pioneer of the CRISPR gene-editing technology, will co-lead one of the project’s four thrusts.

Overall, the project will focus on mathematically grounded machine learning approaches to evaluate, predict, optimize and monitor complex networked systems that are inscrutable to conventional approaches. The key lies in understanding causal relationships in situations where causality is extremely difficult to tease apart, and then using that deep understanding to design interventions.

In looking at the 20,000 genes in a single human cell, for example, the researchers know that these genes help regulate each other in complicated ways but don’t necessarily understand what precise sequences of changes will lead to specific outcomes. Editing these genes to, say, turn a sick cell into a healthy cell while eliminating unintended consequences, is essential to the advancement of personalized medicine.

The team’s principal investigators are: Alberto Abadie, MIT; Miguel Hernan, Harvard; John Ioannidis, Stanford; Devavrat Shah, MIT; Caroline Uhler, MIT; Mengdi Wang, Princeton; and Feng Zhang, Broad Institute.