DrugCLIP

Summary

DrugCLIP embeds protein pockets and small molecules into a shared latent space so large compound libraries can be searched by dense retrieval. It was designed to make genome-wide virtual screening faster than conventional docking and pairwise deep-learning scoring.

Framework

Training combines synthetic protein-ligand data with experimentally determined complexes. GenPack, a generative pocket-refinement component, improves pocket detection on predicted structures, supporting screening against targets modeled with tools such as AlphaFold 2.[1]

The authors reported stronger speed and accuracy than selected docking and deep-learning baselines on DUD-E and LIT-PCBA, followed by wet-lab validation and a public genome-wide screening resource.[1]

Citations

[1] Jia et al. (2026), "Deep contrastive learning enables genome-wide virtual screening"