Assessment of scoring functions for computational models of protein-protein interfaces
Phys. Rev. E 113, 054417 – Published 28 May, 2026
DOI: https://doi.org/10.1103/krhp-fxv3
Abstract
An important goal of computational studies of protein-protein interfaces (PPIs) is to predict the binding site between two monomers that form a heterodimer. The simplest version of this problem is to rigidly redock the bound forms of the monomers, which involves generating computational models of the heterodimer and then scoring them to determine the most nativelike models. PPI scoring functions have been assessed previously using rank- and classification-based metrics; however, these methods are sensitive to the number and quality of models in the scoring function training set. We assess the accuracy of seven physical, statistical, or deep-learning-based PPI scoring functions by comparing their scores of computational models of PPIs to a measure of structural similarity to the x-ray crystal structure (i.e., the DockQ score) for a nonredundant set of heterodimers from the Protein Data Bank. For each heterodimer, we generate redocked models uniformly sampled over DockQ and calculate the Spearman correlation between the PPI scores and DockQ. For some targets, the scores and DockQ are highly correlated; however, for many targets, there are weak correlations. Several physical features explain the difference between difficult- and easy-to-score targets. Strong correlations exist between the score and DockQ for targets with highly intertwined monomers and many interface contacts. We also develop a new score based on only two physical features that matches the performance of current PPI scoring functions. In addition, we address the more general problem of flexible-body docking by generating and docking intermediate monomer conformations between their bound and unbound forms. We score the docked models and find that the Spearman correlations between the PPI scores and DockQ decrease strongly as the monomers are deformed from their bound conformations. These results emphasize that PPI docking predictions can be improved by focusing on correlations between the PPI score and DockQ and incorporating more discriminating physical features into PPI scoring functions.
