One step forward towards deep‐learning protein complex structure prediction by precise multiple sequence alignment construction.

The article discusses the development of an AI-based algorithm called DMFold for predicting the structure of protein-protein complexes. Proteins carry out important biological functions through their interactions with other proteins, known as protein-protein interactions (PPIs). Determining the three-dimensional structures of PPI complexes is crucial for understanding their functions and developing drugs that target PPIs. The DMFold algorithm constructs precise multiple sequence alignments (MSAs) and utilizes co-evolutionary information to predict the structure of PPI complexes. In blind tests, DMFold outperformed other methods, including the widely acclaimed AlphaFold2, in predicting the structures of protein complexes. The algorithm has the potential to enhance protein function annotations and aid in the discovery of drugs targeting PPI-related diseases. [Extracted from the article]