Computational prediction of protein-protein interactions

Area of Science:

• Proteomics
• Computational Biology
• Bioinformatics

Background:

• Eukaryotic proteins feature modular domains and sequence motifs essential for function and post-translational modifications.
• Predicting protein interactions computationally requires identifying these domains and motifs to map their specificities.
• Existing methods struggle with accurately identifying short sequence motifs recognized by protein domains.

Purpose of the Study:

• To develop and present improved computational methods for identifying sequence motifs recognized by protein domains.
• To introduce Scansite, a web-accessible program for predicting protein-protein interactions based on known domain-motif binding data.
• To detail the usage of Scansite for predicting binding partners and identifying proteins containing specific sequence motifs.

Main Methods:

• Combining experimental techniques (peptide libraries, phage display) with computational approaches and sequence databases.
• Developing sequence specificity-based algorithms to link protein domains with their recognized motifs.
• Utilizing Scansite, a program that leverages experimental binding data to predict interactions.

Main Results:

• Scansite successfully predicts interactions between proteins, focusing on domains critical for cell signaling.
• The program can identify potential binding partners for a given protein based on its domains and motifs.
• Scansite can also identify all proteins within a dataset that contain a specific query sequence motif.

Conclusions:

• Improved methods combining experimental and computational data enhance the identification of protein sequence motifs.
• Scansite provides a valuable tool for predicting protein interactions and constructing interaction networks.
• The Scansite program facilitates the discovery of protein binding partners and the identification of proteins harboring specific functional motifs.