Ranking protein-protein models with large language models and graph neural networks

Protein - protein interacnullons (PPIs) are associated with various diseases, including cancer, infecnullons, and neurodegeneranullve disorders. Obtaining three - dimensional structural informanullon on these PPIs serves as a foundanullon to interfere with those or to guid e drug design. Various strategies can be followed to model those complexes, all typically resulnullng in a large number of models. A challenging st e p in this process is the idennullfica-nullon of good models ( near - nanullve PPI conformanullons) from the large pool of generated models . T o address this challenge, we previously developed DeepRank - GNN - esm, a graph - based deep learning algorithm for ranking modelled PP I structures harnessing the power of protein language model s .