zPocket: open cofolding-feature pocket predictor hits 0.745 AUPRC, beats P2Rank
LotfiSanae · x · 2026-10-15
Nate Gruver, Tally Portnoi et al. released zPocket, a protein-ligand binding pocket predictor built on frozen co-folding model features, matching closed IsoDDE and AnewDDE performance while fully reproducible and open when using ESMFold2.
The paper also introduces PocketWatch, a curated benchmark of 1,275 test proteins stratified by sequence similarity, structural similarity, and crypticity, fixing the issue of existing benchmarks falling inside co-folding models' training distribution. The best zPocket model achieves 0.745 overall per-target AUPRC vs 0.598 (VN-EGNN) and 0.595 (UniSite-1D) retrained on the split — roughly 1.5x P2Rank's 0.477. It also improves top-N+2 recall via per-residue clustering or rescoring centroids from other models. Paper: arxiv.org/abs/2610.17696
More from Research
- RAI Institute's single-track bicycle robot UMV lands in Science Robotics — joemeno · 2026-10-15
- Five-year Harvard postdoc preprint unveils BARRACUDA for systematic synthetic circuit design — NikoMcCarty · 2026-10-15
- Autonomous LLM agents for materials science: seminar at Texas Southern University — Fowe · 2026-10-15
- Behavioral Self-Rewarding lifts Qwen3-14B to 34.7% on SWE-bench Verified — dair_ai · 2026-10-15
- vLLM adaptive verification now works head-free, using draft logits with online logistic calibration — vllm_project · 2026-10-15
- New arXiv paper: thousands of ML models poorly predict hundreds of life outcomes — RobbWiller · 2026-10-15