Deep-learning in protein-protein interactions identifies complexes that will advance our understanding of cellular processes

Using a combination of RoseTTAFold and AlphaFold to screen millions of pairs of yeast proteins, researchers identified over 1,500 pairs likely to interact. These complexes, which had as many as 5 subunits, play roles in almost all key processes in eukaryotic cells and provide broad insights into biological function. “…our results herald a new era of structural biology in which computation plays a fundamental role in both interaction discovery and structure determination,” write the authors. Protein-protein interactions play critical roles in biology, but the structures of many eukaryotic protein complexes are unknown, and there are likely many interactions not yet identified. Recent deep-learning-based advances in protein structure prediction, including those highlighted in a 2021 study in Science that presented the tool RoseTTAFold, have the potential to increase the power of approaches for identifying pairs of interacting proteins. Ian Humphreys and colleagues took advantage of such recent advances. They combined RoseTTAFold and DeepMind’s AlphaFold to screen through 8.3 million pairs of yeast proteins, identifying 1,505 likely to interact. They built structure models for 106 previously unidentified assemblies and for 806 that have not been structurally characterized. “Our approach extends the range of large scale deep learning based structure modeling from monomeric proteins to protein assemblies,” the authors say. “…following up on the many new complexes presented here should advance understanding of a wide range of eukaryotic cellular processes and provide new targets for therapeutic intervention.”