Interactome INSIDER

Co-crystal structures are the most reliable source of partner-specific interface residue information. We compile all available co-crystal structures from the RCSB Protein Data Bank, resolving the interface of 7,135 interactions through their 3D structures.

Homology modelling allows us to infer a likely configuration for protein-protein interaction using known homologous structures. We expand our interface residue repository by incorporating homology models via Interactome3D for interactions with no known co-crystal, adding 5,386 interaction interfaces.

We train ECLAIR (Ensemble Classifier Learning Algorithm to predict Interface Residues) to predict partner-specific interaction interfaces using biophysical, conservation, coevolution, structural, and docking features without requiring they all be available. Using ECLAIR, we predict interface residues for all 184,605 remaining interactions.

We compile a collection of more than 56,000 disease-associated mutations from both HGMD and ClinVar, additionally processing more than 1.3 million somatic mutations from COSMIC. These mutations can be explored in the context of interface domains, residues, or 3D atomic clustering, providing multi-scale information for genomic studies.

Users can submit their own set of mutations to explore enrichment at different protein-protein interaction interfaces. Users who submit their own set will receive a Mutation Code, which they can use to return to the site and explore interface enrichment at a later time.

We compile more than 600,000 Population variants from all populations contained in the 1000 Genomes Project, ESP, and UniProt. Users can explore selective pressure in the context of both interface residues and disease mutations.

We provide deleteriousness predictions for each disease mutation by leveraging PolyPhen-2, SIFT, and the Grantham Score of each amino acid substitution in our databases. These predictors are also available for user-submitted mutation sets.

Biophysical changes arising from a substitution and its location relative to the surface of the protein are key pieces of information for functional inference. We presented these in addition to Jensen Shannon Divergence, a measurement of amino acid conservation, to further contextualize mutations as they relate to interaction.

We provide interactions from a curated compilation of high-quality binary protein-protein interactions from 7 interactome resources (BioGRID, MINT, iRefWeb, DIP, IntAct, HPRD and, MIPS). Interactions are filtered both systematically and manually to remove erroneous and low-quality interactions.

Overall we provide interface information for 121,575 experimentally-determined human binary interactions reported in major databases. We also provide downloads for interface information for high-quality binary interactomes in 7 highly studied model organisms: A. thaliana, S. cerevisiae, C. elegans, D. melanogaster, M. musculus, S. pombe, and E. coli.