ProMateus

ProMateus predicts and characterizes protein–protein interaction interfaces from protein structural features to identify binding-site locations and their biophysical determinants.

Key Features:

• Surface analysis: Probes protein surfaces using spheres of 10 Å radius and analyzes a database of 57 unique, non-homologous proteins with heteromeric, transient protein–protein interactions to characterize binding-site preferences.
• Structural preferences: Reports enrichment of β-sheets and relatively long non-structured chains at interfaces, no corresponding enrichment for α-helices, and a chemical preference for aromatic side chains at interfaces.
• Chemical characteristics: Identifies clustering of hydrophobic and polar residues at interfaces despite similar overall hydrophobic/polar composition to the remainder of the surface, and reports more bound water molecules and lower B-factors at binding sites even in the unbound state.
• Prediction algorithm: Incorporates interface-derived properties into an automated predictor that reports approximately 70% success in identifying binding-site locations on unbound proteins with known structures and shows consistent performance on unbound and bound datasets.
• Community-driven feature integration: Supports incorporation and testing of user-suggested features and parameters inspired by existing predictors (e.g., WHISCY, PPI-Pred, PINUP).
• Framework for feature exploration: Enables systematic evaluation of individual features and feature combinations to compare and optimize predictive contributions across methods.

Scientific Applications:

• Interaction-site mapping: Localizes likely protein–protein interaction sites on known structures and characterizes determinants of binding specificity and kinetics.
• Proteome-scale prioritization: Supports proteome-scale analyses to prioritize candidate interfaces for experimental validation.
• Method development and feature discovery: Provides an extensible framework for integrating and comparatively evaluating new predictive signals for interface prediction.

Methodology:

Extends the ProMate framework to incorporate new features and additional datasets within a unified prediction and evaluation pipeline; optimization and assessment are performed with procedures that exclude the proteins under test from feature derivation and data collection to reduce overfitting, and the platform incorporates and tests parameters inspired by existing predictors (e.g., WHISCY, PPI-Pred, PINUP).