GRAMM-X

GRAMM-X predicts and ranks protein–protein docking poses to support structural analysis of protein–protein interactions.

Key Features:

• Smoothed Potentials: Employs smoothed potential functions to accommodate conformational changes during docking and improve binding-interface predictions.
• Refinement Stage: Refines top poses using conjugate gradient minimization with smoothed potentials to optimize local minima from global searches.
• Knowledge-Based Scoring: Integrates knowledge-based scoring that includes pairwise residue–residue statistical preferences, cluster occupancy, and evolutionary conservation of interfaces for solution ranking.
• FFT-based Global Search: Extends the GRAMM Fast Fourier Transformation approach by using a projection of a smoothed Lennard-Jones potential on a fine grid for initial matching.
• High-Performance Computing: Processes docking problems on a 320-processor Linux cluster for computationally intensive searches.

Scientific Applications:

• Binding-site prediction: Predicts initial binding sites of protein–protein complexes for structural biology studies.
• Low-resolution docking: Performs low-resolution docking suitable when high-resolution data are unavailable and has been applied to predict an influenza virus hemagglutinin–antibody complex.
• Benchmarking and validation: Has been benchmarked through public use and CAPRI participation (including CAPRI Round 5) with predictions assessed by RMSD.

Methodology:

Two-stage docking: a global fine-grid search using a projection of a smoothed Lennard-Jones potential to identify matches, followed by local optimization via conjugate gradient minimization of top predictions and reranking by a composite scoring system including Lennard-Jones terms and knowledge-based scores (pairwise residue–residue preferences, cluster occupancy, evolutionary conservation).