RaptorX-ComplexContact

RaptorX-ComplexContact predicts residue-residue contacts between interacting proteins by integrating sequential and coevolutionary signals from paired multiple sequence alignments to inform protein docking and interaction analysis.

Key Features:

• Deep Learning Framework: Employs two deep convolutional residual neural networks (ResNet) to extract patterns from sequential features and coevolution information derived from paired MSAs.
• Paired Multiple Sequence Alignments (MSA): Generates paired MSAs for a given pair of interacting protein sequences to capture coevolutionary signals.
• Genomic Distance-Based Alignment: Produces MSAs based on genomic proximity to reflect evolutionary relationships between interacting genes.
• Phylogeny-Informed Alignment: Incorporates phylogenetic information to refine paired MSAs and improve detection of coevolutionary signals.
• DCA Limitation Mitigation: Addresses limitations of direct-coupling analysis (DCA) that require extensive sequence homologs by integrating multiple signal types via deep learning.
• Interprotein Contact Prediction: Integrates sequential and coevolutionary data through the deep learning models to predict interprotein residue-residue contacts.

Scientific Applications:

• Protein Docking: Provides predicted residue-residue interactions to inform modeling of protein-protein docking interfaces.
• Protein-Protein Interaction Prediction: Facilitates identification and characterization of potential interacting protein partners through contact inference.
• Protein Interaction Network Construction: Supplies contact-derived evidence for building maps of protein interactions within cells.

Methodology:

RaptorX-ComplexContact generates paired MSAs using genomic-distance and phylogeny-informed approaches, extracts sequential and coevolutionary features from the MSAs, and uses two deep convolutional residual neural networks (ResNet) that integrate these inputs to predict interprotein residue-residue contacts.