Abstract
AlphaFold-Multimer models protein complexes and facilitates protein-protein interaction (PPI) prediction. Mapping of protein interaction domains and motifs onto the 3D structure can lend credence to the model and provide insight into the function of a given interaction. Furthermore, limiting structure prediction to only the domains and motifs that are likely to interact can reduce the computational demand and produce a higher quality model. To satisfy these needs, we built the Protein-Protein Interaction Identifier (PPI-ID). PPI-ID maps interaction domains and motifs onto molecular structures and filters for those that are sufficiently close to interact. Once an interface is found, PPI-ID labels interacting amino acids. Given only sequences, PPI-ID predicts regions for AlphaFold-Multimer modeling, reporting potential interactions only when each protein has one-half of a paired sequence. Testing with known dimers confirms high accuracy of the tool.