Protein Interactions Predicted by Artificial Intelligence

Researchers from UT Southwestern and the University of Washington led an international team that created 3D models of eukaryotic Protein interactions using artificial intelligence (AI) and evolutionary analysis. For the first time, the study published in Science identified over 100 likely Protein complexes and gave structural models for over 700 previously uncharacterized ones. Understanding how Protein pairs or groups interact to carry out biological functions could lead to a slew of new therapeutic targets.

Dr. Cong co-led the work with David Baker, Ph.D., Professor of Biochemistry at the University of Washington, and Dr. Cong’s postdoctoral adviser before coming to UT Southwestern. Jimin Pei, Ph.D., a computational biologist at UT Southwestern, is one of the study’s four co-lead authors. Dr. Cong noted that Protein frequently work in pairs or groups known as complexes to complete all of the tasks required to keep an organism alive. While some of these connections have been thoroughly researched, others remain a mystery.

Building entire interactomes – or representations of a cell’s whole network of molecular interactions – would shed light on many fundamental elements of biology and provide researchers with a new starting point for developing medications that promote or discourage these connections. Dr. Cong is a researcher in interactomics, a new discipline that blends bioinformatics and biology. Uncertainty about the structures of many was a key hurdle to developing an interactome until recently, a problem that scientists have been seeking to tackle for half a century.

Dr. Cong, Dr. Baker, and his colleagues built on those AI structure-prediction methods in the current study by modeling a variety of yeast Protein complexes. Yeast is a popular model organism for basic biological research. To locate that were likely to interact, the researchers looked through the genomes of related fungus for genes that had undergone similar changes.