Accelerate Protein Docking with EquiDock on Vecura

What is EquiDock?

EquiDock is an SE(3)-equivariant geometric deep-learning model designed for end-to-end rigid-body protein-protein docking. By encoding proteins as residue-level geometric graphs and utilizing an Invariant Equivariant Graph Neural Network (IEGNN), it predicts the precise rotation and translation required to dock a ligand protein onto a receptor in a single forward pass. This approach replaces traditional, time-consuming candidate-pose sampling with a highly efficient, symmetry-aware architectural design.

It helps users rapidly generate docked protein-protein complexes directly from their unbound structures. It is especially useful for structural biologists and drug discovery researchers who require fast, high-throughput modeling of protein interactions without the overhead of massive computational docking simulations.

What can users do with EquiDock on Vecura?

With EquiDock on Vecura, users can:

• Generate 3D docked complex models by uploading unbound receptor and ligand PDB files.
• Automatically compute the necessary SE(3) rigid-body transformations.
• Apply optional, automated post-processing to minimize steric clashes between protein partners.
• Produce standardized output files that are immediately ready for downstream visualization, energy analysis, or flexible refinement.

What the output means

The output provides the predicted 3D structure of the docked complex in PDB format, the transformed coordinates of the ligand, and the calculated rigid-body transformation (rotation matrix and translation vector).

This output should be used to support scientific decision making. It does not replace experimental validation.

Why this matters

Protein-protein interactions (PPIs) are central to almost all biological processes, and understanding their structural interface is a cornerstone of drug discovery and synthetic biology. Traditional docking methods often rely on sampling thousands of potential configurations, which is computationally expensive and slow. EquiDock’s ability to predict a single, accurate binding pose in seconds significantly accelerates the early stages of structural analysis.

By removing the need for extensive sampling and complex energy-function refinement, EquiDock lowers the barrier to entry for performing initial docking experiments, enabling researchers to explore larger interaction spaces more efficiently.

• Developed by: Octavian Ganea et al.
• Source: ICLR 2022 Paper / GitHub Repository
• Reference: EquiDock: Geometric Deep Learning for Rigid-Body Protein-Protein Docking