Residue co-evolution helps predict interaction sites in α-helical membrane proteins

Area of Science:

• Computational biology
• Structural biology
• Bioinformatics

Background:

• Integral membrane proteins form multi-subunit complexes for cellular functions.
• Experimental determination of protein complex structures is limited.
• Computational prediction of interaction interfaces is vital for functional annotation.

Purpose of the Study:

• To develop a sequence-based method (MBPred) for predicting interface residues in transmembrane proteins.
• To address the challenge of low structural coverage for membrane protein complexes.

Main Methods:

• Developed MBPred, a sequence-based prediction method using random forest models.
• Implemented separate models for cases with known and predicted transmembrane region locations.
• Utilized residue co-evolution as an informative feature.

Main Results:

• MBPred achieved high performance metrics (AUC 0.79, precision 0.69, recall 0.55 on cross-validation).
• The method demonstrated superior performance compared to existing approaches for both transmembrane and globular proteins.
• MBPred successfully identified over 50% of residues in most complete interface patches.

Conclusions:

• MBPred is an effective tool for predicting transmembrane protein interaction sites.
• The method's performance highlights the utility of residue co-evolution in interface prediction.
• MBPred advances the functional annotation of membrane proteins.