This work introduces Practical Molecular Optimization for Docking (PMO-Dock), a benchmark and protocol for docking-based optimization consisting of 25 tasks covering hit generation, lead optimization, and a new specificity task requiring strong on-target binding while penalizing off-target interactions.
Abstract
The Practical Molecular Optimization (PMO) benchmark standardized evaluation in molecular optimization, but it is built on simple property-based oracles that do not capture structure-based drug design. As the field has moved to docking-based objectives, evaluation practice has become unstandardized, and different methods report results on different docking tasks, often without a shared protocol for fair comparison or sample-efficiency constraints. To address this, we introduce Practical Molecular Optimization for Docking (PMO-Dock), a benchmark and protocol for docking-based optimization consisting of 25 tasks covering hit generation, lead optimization, and a new specificity task requiring strong on-target binding while penalizing off-target interactions. The protocol separates model development from final evaluation: validation tasks are used for hyperparameter tuning, and test tasks are held out for final evaluation and are not used during development. Evaluation on the test tasks is done under a strict oracle budget to reflect realistic drug-discovery settings. We benchmark four diverse high-performing methods spanning different optimization paradigms, Saturn (reinforcement learning), GenMol (discrete diffusion), Genetic-guided GFlowNet, and ChemLlama (LLM-based evolutionary search). Our results show no universally best method across tasks, additional analyses of hyperparameter transfer (cross-task and cross-category) and sensitivity show that transfer is often limited and varies by method, motivating method-dependent selection strategies. This benchmark provides a concrete standard for evaluating generalization to unseen tasks and sample efficiency under a limited test-time oracle budget.
Ligand–protein interaction prediction is essential for structure-based drug discovery, but further development of machine-learning scoring functions is limited by a lack of high-quality training data sets. Here, we present DockBind, a curated data set of modeled protein–ligand complexes paired with experimental binding...
Alberto Rovetta, Alberto Memè, H. Antila et al.· Journal of Chemical Informat...· 0 citations
Joint molecular optimization serves as the central focus of this research, which explores the synergistic integration of machine learning (ML) and virtual screening (VS) for drug discovery. This study addresses these limitations by integrating molecular docking and ML to predict and balance multiple molecular propertie...
Machine learning-based quantitative structure–activity relationship (ML-QSAR) modeling requires consistent data preparation, validation, and model comparison. We present LASSBio-classFLOW, an open, modular, and semiautomated KNIME workflow for classification-based ligand screening. It integrates basic molecular structu...
Pedro de Sena Murteira Pinheiro, Jefferson Muniz Alves da Silva, Bárbara da Silva Mascarenhas de Jesus et al.· ACS Omega· 0 citations
Virtual screening answers two different questions with one number: enrichment, separating actives from background molecules, and potency ranking, ordering actives by measured activity. We benchmarked both for the same score on the same compounds across five CNS binding sites (SERT, GABAA benzodiazepine site, 5-HT1A, σ1...
Y. Lai, Hsiang-Chi Huang· International Journal of Mol...· 0 citations
Predicting how a ligand binds to its target is central to small-molecule drug discovery. Modern cofolding models can generate diverse candidate poses, but reliably ranking these candidates remains challenging. We introduce Pyrite, a pairwise pose-ranking benchmark comprising 1,500 pose pairs across diverse protein-liga...
Megha Srivastava, Michael H. Lim, Allison M. Keys et al.· 1 citation
Rescoring docked ligands with molecular dynamics (MD) and end point free energy methods such as MM/GBSA is widely used, but its benefit over docking for series of close analogs is uncertain. We compared AutoDock Vina scores with Md based MM/GBSA energies for 29 neutral p38α kinase inhibitors (PDB 3FLY) from the Open Fo...
By studying the computational foundations of language learning, Associate Professor Jacob Andreas builds intelligent systems that can communicate effectively with humans.
Microsoft Research Blog· microsoft.comNov 16, 2026
AI coding agents are increasingly help developers fix bugs and build features largely on their own, but every interaction with their models consumes tokens. In each session, an agent sends instructions, tool descriptions, and session history to its model. Most efforts to reduce this cost focus on individual tasks. Yet developers return to the same […] The post TokenEconomizer: Learning from Experience to Make AI Coding Agents More Efficient appeared first on Microsoft Research.
MIT News · Artificial Intelligence· news.mit.eduNov 13, 2026