Predicting the\nDynamic Interaction of Intrinsically\nDisordered Proteins

Intrinsically disordered proteins (IDPs) participate\nin various\nbiological processes. Interactions involving IDPs are usually dynamic\nand are affected by their inherent conformation fluctuations. Comprehensive\ncharacterization of these interactions based on current techniques\nis challenging. Here, we present GSALIDP, a GraphSAGE-embedded LSTM\nnetwork, to capture the dynamic nature of IDP-involved interactions\nand predict their behaviors. This framework models multiple conformations\nof IDP as a dynamic graph, which can effectively describe the fluctuation\nof its flexible conformation. The dynamic interaction between IDPs\nis studied, and the data sets of IDP conformations and their interactions\nare obtained through atomistic molecular dynamic (MD) simulations.\nResidues of IDP are encoded through a series of features including\ntheir frustration. GSALIDP can effectively predict the interaction\nsites of IDP and the contact residue pairs between IDPs. Its performance\nin predicting IDP interactions is on par with or even better than\nthe conventional models in predicting the interaction of structural\nproteins. To the best of our knowledge, this is the first model to\nextend the protein interaction prediction to IDP-involved interactions.