Prediction of interacting proteins from homology-modeled complex structures using sequence and structure scores

Area of Science:

• Computational biology
• Biochemistry
• Bioinformatics

Background:

• Protein-protein interactions (PPIs) are fundamental to cellular processes.
• Experimental methods for PPI identification generate vast data but have limitations in scope and accuracy.
• Computational prediction of PPIs from sequence or structure is valuable for elucidating cellular functions.

Purpose of the Study:

• To develop and evaluate a computational method for predicting protein-protein interactions using homology-modeled complex structures.
• To assess the contribution of sequence similarity and structural scores in predicting interacting proteins.

Main Methods:

• Utilized statistical residue-residue contact energy, simple electrostatic energy, and sequence similarity scores.
• Modeled protein-protein complex structures for *Saccharomyces cerevisiae* heterodimers.
• Evaluated prediction performance using a complete dataset (10,325 models) and a high-confidence dataset (3,219 models).

Main Results:

• Sequence similarity demonstrated significantly higher discrimination power than structure-based scores.
• Incorporating contact energy improved predictions beyond using sequence similarity alone.
• Both complete and high-confidence datasets supported the findings.

Conclusions:

• Sequence similarity is an essential component for predicting protein-protein interactions.
• Structure-based scores, while less powerful individually, enhance prediction accuracy when combined with sequence similarity.
• The developed method offers a valuable computational approach for identifying interacting protein partners.