maabuu/posebusters

Plausibility checks for generated molecule poses.

What it solves

PoseBusters provides a way to verify if generated molecule poses (such as those produced by AI-based docking methods) are physically plausible. It addresses the problem where AI models may generate poses that look correct but are physically invalid, failing to generalise to novel sequences.

How it works

The tool performs plausibility checks on molecule poses provided in SDF format. It can evaluate a single molecule, compare a predicted pose against a true crystal structure, or check a ligand generated for a specific protein (PDB format).

Who it’s for

Computational chemists and researchers working on molecule docking and AI-based drug discovery who need to validate the predicted 3D structures of ligands.

Highlights

  • Plausibility checks for generated molecule poses.
  • Support for checking single molecules, protein-ligand complexes, and re-docked ligands.
  • Command-line interface for batch processing via CSV files.
  • Based on research published in Chemical Science.

Related

  • Project
  • Project
  • Project
  • Project