diffdock

SKILL.md

DiffDock is a diffusion-based deep learning tool for molecular docking that predicts 3D binding poses of small molecule ligands to protein targets. It represents the state-of-the-art in computational docking, crucial for structure-based drug discovery and chemical biology.

Key Distinction: DiffDock predicts binding poses (3D structure) and confidence (prediction certainty), NOT binding affinity (ΔG, Kd). Always combine with scoring functions (GNINA, MM/GBSA) for affinity assessment.

This script validates Python version, PyTorch with CUDA, PyTorch Geometric, RDKit, ESM, and other dependencies.

Diffusionsbasiertes molekulares Andocken. Prognostizieren Sie Protein-Ligand-Bindungspositionen anhand von PDB/SMILES, Konfidenzwerten und virtuellem Screening für strukturbasiertes Arzneimitteldesign. Nicht für Affinitätsvorhersage. Quelle: ovachiever/droid-tings.

Original anzeigen