Breaking Bad Molecules: Are MLLMs Ready for Structure-Level Molecular Detoxification?

2025-06-12

AI Summary

Molecular toxicity repair—generating structurally valid molecules with significantly reduced toxicity—has long lacked a formal task definition and standardized benchmarking. This work formally defines the task for the first time and introduces ToxiMol, the first MLLM-oriented benchmark comprising 560 mechanistically diverse toxic molecules and 11 structural editing tasks. We propose a mechanism-aware prompting annotation paradigm and ToxiEval, an end-to-end automated evaluation framework integrating toxicity prediction, synthetic accessibility, drug-likeness (QED), and structural similarity. Systematic evaluation of nearly 30 state-of-the-art MLLMs reveals their nascent capabilities in toxicity comprehension and structural editing, while uncovering critical bottlenecks—including insufficient candidate diversity and mechanism attribution bias. ToxiMol and ToxiEval establish foundational resources and methodologies for advancing MLLM-driven molecular toxicity mitigation.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: Safety and Robustness
• Philosophy and Ethics of AI: Safety, Robustness & Trustworthiness

Application Category

• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

Toxicity remains a leading cause of early-stage drug development failure. Despite advances in molecular design and property prediction, the task of molecular toxicity repair - generating structurally valid molecular alternatives with reduced toxicity - has not yet been systematically defined or benchmarked. To fill this gap, we introduce ToxiMol, the first benchmark task for general-purpose Multimodal Large Language Models (MLLMs) focused on molecular toxicity repair. We construct a standardized dataset covering 11 primary tasks and 560 representative toxic molecules spanning diverse mechanisms and granularities. We design a prompt annotation pipeline with mechanism-aware and task-adaptive capabilities, informed by expert toxicological knowledge. In parallel, we propose an automated evaluation framework, ToxiEval, which integrates toxicity endpoint prediction, synthetic accessibility, drug-likeness, and structural similarity into a high-throughput evaluation chain for repair success. We systematically assess nearly 30 mainstream general-purpose MLLMs and design multiple ablation studies to analyze key factors such as evaluation criteria, candidate diversity, and failure attribution. Experimental results show that although current MLLMs still face significant challenges on this task, they begin to demonstrate promising capabilities in toxicity understanding, semantic constraint adherence, and structure-aware molecule editing.

Problem

• Benchmarking MLLMs for molecular toxicity repair
• Standardizing toxicity repair tasks across diverse molecules
• Evaluating MLLM performance in structure-aware detoxification

Innovation

• ToxiMol benchmark for MLLM toxicity repair
• Prompt annotation with expert toxicological knowledge
• ToxiEval automated multi-criteria evaluation framework