SelectBench: first selectivity benchmark finds generative SBDD models fall short
dom_beaini · x · 2026-10-24
arXiv paper 2610.24124 introduces SELECTBENCH, the first benchmark evaluating selectivity of generative structure-based drug design (SBDD) models — molecules binding the right protein can still cause serious side effects by binding off-target proteins. It offers three tasks (literature case studies, pharma safety screening panels, large-scale structure-based sets) plus metrics to quantify selectivity. Evaluating nine representative SBDD models, including selectivity-trained methods and inference-time guidance, the authors find only modest off-target discrimination and limited guidance gains, underscoring the need for better methods.
More from Research
- Economists debate whether AI-era economics will produce 10x new economists or just more empiricism — soumitrashukla9 · 2026-10-25
- EMNLP Paper: Repetition Mismatch Explains Why Data Mixture Experiments Don't Scale — MarekRei · 2026-10-25
- Imperial NLP Brings 6 Main Papers and Agent-LM Workshop to EMNLP 2026 — MarekRei · 2026-10-25
- Google's Jon Barron shares his full AI paper-writing workflow with Gemini — lmoroney · 2026-10-25
- Only 30% of suspected rare-disease newborns get diagnosed; Gamow Labs launches 122-case RareBench for agentic AI — Sauers_ · 2026-10-25
- MMGR benchmark: top video models solve only ~3% of mazes despite stunning visuals — zhaofeng_wu · 2026-10-25