FORGE: Fragment-Oriented Ranking and Generation for Context-Aware Molecular Optimization
2026-05-11 • Machine Learning
Machine Learning
AI summaryⓘ
The authors present FORGE, a new method to improve molecules by making smart, small changes based on their context. Instead of using natural language, which can be inaccurate and limited by data size, FORGE uses pairs of molecule fragments to learn what changes help. It works in two steps: first, it finds important fragments to change, then it suggests replacements for those fragments. This approach outperforms bigger models and other techniques while avoiding errors caused by language-based methods.
molecular optimizationlanguage modelsfragment editingcontext-awarechemical propertiessequence generationblack-box objectivesmachine learningPrompt-MolOptgraph methods
Authors
Qingchuan Zhang, He Cao, Hao Li, Yanjun Shao, Zhiyuan Liu, Shihang Wang, Shufang Xie, Shenghua Gao, Xinwu Ye
Abstract
Molecular optimization seeks to improve a molecule through small structural edits while preserving similarity to the starting compound. Recent language-model approaches typically treat this task as prompt-conditioned sequence generation. However, relying on natural language introduces an inherent data-scaling bottleneck, often leads to chemical hallucinations, and ignores the strong context dependence of fragment effects. We present FORGE, a two-stage framework that reformulates molecular optimization as context-aware local editing. By utilizing automatically mined, verified low-to-high edit pairs instead of expensive human text annotations, Stage 1 ranks candidate fragments by their property contribution under the full molecular context to inject chemical prior, and Stage 2 generates explicit fragment replacements. Built on a compact 0.6B language model, FORGE further adapts to unseen black-box objectives through in-context demonstrations. Across Prompt-MolOpt, PMO-1k and ChemCoTBench, FORGE consistently outperforms prior methods, including substantially larger language models and graph methods. These results highlight the value of explicit fragment-level supervision as a more easily obtainable, scalable, and hallucination-less alternative to natural language training.