Conditional Chemical Language Models are Versatile Tools in Drug Discovery

2025-07-14

AI Summary

Generative chemical language models (CLMs) face two key bottlenecks in drug discovery: unreliable reward signals and poor output interpretability. To address these, we propose SAFE-T—a fragment-level, biologically contextualized CLM framework conditioned on target proteins and mechanisms of action. SAFE-T unifies virtual screening, target prediction, and activity cliff detection in a single multitask architecture—without requiring 3D structural information or handcrafted scoring functions. Its core innovations are: (1) zero-shot transfer via conditional fragment-sequence modeling; (2) fragment attribution for interpretable structure–activity relationship analysis; and (3) tight integration of scoring and generation to accelerate inference. SAFE-T achieves state-of-the-art performance across diverse prediction and generation benchmarks, demonstrating both high efficiency and strong generalization in early-stage drug discovery.

Technology Category

• Natural Language Processing: Safety and Robustness
• Machine Learning: Large Multimodal Models (LMMs)
• Constraint Satisfaction and Optimization: Satisfiability Modulo Theories

Application Category

• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

Generative chemical language models (CLMs) have demonstrated strong capabilities in molecular design, yet their impact in drug discovery remains limited by the absence of reliable reward signals and the lack of interpretability in their outputs. We present SAFE-T, a generalist chemical modeling framework that conditions on biological context -- such as protein targets or mechanisms of action -- to prioritize and design molecules without relying on structural information or engineered scoring functions. SAFE-T models the conditional likelihood of fragment-based molecular sequences given a biological prompt, enabling principled scoring of molecules across tasks such as virtual screening, drug-target interaction prediction, and activity cliff detection. Moreover, it supports goal-directed generation by sampling from this learned distribution, aligning molecular design with biological objectives. In comprehensive zero-shot evaluations across predictive (LIT-PCBA, DAVIS, KIBA, ACNet) and generative (DRUG, PMO) benchmarks, SAFE-T consistently achieves performance comparable to or better than existing approaches while being significantly faster. Fragment-level attribution further reveals that SAFE-T captures known structure-activity relationships, supporting interpretable and biologically grounded design. Together with its computational efficiency, these results demonstrate that conditional generative CLMs can unify scoring and generation to accelerate early-stage drug discovery.

Problem

• Overcoming unreliable reward signals in generative chemical language models
• Enhancing interpretability in molecular design outputs
• Unifying scoring and generation for early-stage drug discovery

Innovation

• Conditions on biological context for molecular design
• Models fragment-based sequences with biological prompts
• Enables interpretable and efficient drug discovery