Breaking Bad Molecules: Are MLLMs Ready for Structure-Level Molecular Detoxification?
Authors: Fei Lin, Ziyang Gong, Cong Wang, Tengchao Zhang, Yonglin Tian, Yining Jiang, Ji Dai, Chao Guo, Xiaotong Yu, Xue Yang, Gen Luo, Fei-Yue Wang
Organizations: Department of Engineering Science, Macau University of Science and Technology, Macau, China · School of Computer Science, Shanghai Jiao Tong University, Shanghai, China · Institute of Automation, Chinese Academy of Sciences, Beijing, China · School of Pharmacy, Macau University of Science and Technology, Macau, China · Faculty of Electrical Engineering and Computer Science, Ningbo University, Ningbo, China · State Key Laboratory of Biopharmaceutical Preparation and Delivery, Institute of Process Engineering, Chinese Academy of Sciences, Beijing, China · School of Automation and Intelligent Sensing, Shanghai Jiao Tong University, Shanghai, China · Shanghai Artificial Intelligence Laboratory, Shanghai, China
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 660 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 43 mainstream general-purpose MLLMs and conduct multiple ablation studies to analyze key issues, including evaluation metrics, 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 editing.
Figures & tables
Figure 1: Success rates (%) of representative MLLMs [ 1 , 2 , 3 , 4 ] on 11 toxicity repair tasks. Highlighted tasks use a 0–8% scale.
Figure 2: Statistics of toxic molecules and the number of clusters for each task. Values are plotted along the radius on a logarithmic scale ( 101 – 104 ); the dashed ring at n=60 indicates the balanced sampling size determined by the Carcinogens bottleneck constraint.
Figure 3: (a) Global chemical space coverage. UMAP embeddings are computed from ECFP4 fingerprints, where gray points denote all toxic molecules in the source datasets and colored points denote the 660 representative molecules selected via cluster-based sampling (colored by task). (b) A task-level example of cluster-based sampling ( DILI ). Light-blue points denote all 236 toxic molecules in the task, and red points denote the selected 60 representative molecules; the dispersed distribution reflects the diversity-oriented sampling strategy induced by Butina clustering.
Figure 4: A repair example on the AMES task (binary classification, Ssafe=1 ) using the ToxiMol benchmark. The input includes the SMILES of Coumarin 6H [ 92 ] , its 2D molecular image, and a task-adaptive prompt annotation pipeline. After MLLM-mediated toxicity repair, three candidate molecules are generated. Due to structural invalidity, the ToxiEval evaluation chain first filters out Candidate 2. Among the remaining molecules, Candidate 1 is excluded because it fails the safety threshold in the multi-criteria assessment. Candidate 3 passes all stages of the ToxiEval chain and is identified as a successful repair.
Models
LD50
hERG C
Carc L
Tox21
AMES
ToxCast
ClinTox
DILI
hERG K
hERG
SkinRxn
Overall
Domain-Specific MLLMs
ChemVLM 26B [ 67 ]
0.0
26.7
16.7
31.7
1.7
28.3
33.3
1.7
13.3
6.7
8.3
15.3
ChemVLM 8B [ 67 ]
0.0
35.0
20.0
38.3
18.3
43.3
35.0
3.3
16.7
8.3
10.0
20.8
Intern-S1 [ 98 ]
0.0
68.3
43.3
53.3
26.7
48.3
50.0
3.3
40.0
20.0
25.0
34.4
Intern-S1 Mini [ 98 ]
0.0
70.0
35.0
45.0
23.3
55.0
45.0
3.3
23.3
16.7
23.3
30.9
Average Success Rate
0.0
50.0
28.8
42.1
17.5
43.7
40.8
2.9
23.3
12.9
16.7
25.4
Table 1: Toxicity Repair Success Rates (%). hERG C , hERG K , and Carc L refer to the hERG_Central , hERG_Karim , and Carcinogens tasks, respectively. SkinRxn denotes the SkinReaction task.
Figure 6
Task
Type-T (%)
Type-O (%)
Dominant Bottleneck
LD50
43.3
0.0
Dose-toxicity regression bottleneck
hERG C
38.3
1.7
Type-T dominant; Type-O marginal
Carc L
0.0
18.3
Type-O dominant
Tox21
3.3
23.3
Type-O dominant
AMES
11.7
13.3
Mixed bottlenecks (Type-T ≈ Type-O)
ToxCast
0.0
18.3
Type-O dominant
Table 2: Failure-mode attribution for Claude-3.7 Sonnet on each toxicity-repair task. Percentages are computed over all generated candidate molecules for each task; Type-T/Type-O are assigned only to RDKit-valid candidates (others are not attributed to either type).
Closed-Source MLLMs
Open-Source MLLMs
Task
GPT-4.1
GPT-5.1
Qwen3-VL 30B-A3B Instruct
Qwen3-VL 235B-A22B Instruct
LD50
0.3 ± 0.7
0.0 ± 0.0
0.0 ± 0.0
1.3 ± 0.7
hERG C
61.0 ± 4.7
58.7 ± 3.7
30.7 ± 1.7
56.0 ± 1.3
Carc L
34.0 ± 1.5
31.7 ± 3.2
29.3 ± 0.8
30.3 ± 0.7
Tox21
45.0 ± 1.8
44.3 ± 0.8
37.0 ± 1.9
48.7 ± 1.3
AMES
20.0 ± 1.1
23.3 ± 1.1
17.7 ± 1.7
22.7 ± 1.4
Table 4: Multi-round toxicity repair success rates (%) across representative MLLMs (mean ± std over five runs). We report per-task and overall success rates for two closed-source models, GPT-4.1 [ 108 ] and GPT-5.1 [ 107 ] , and two open-source models from the Qwen3-VL family, Qwen3-VL 30B-A3B Instruct [ 106 ] and Qwen3-VL 235B-A22B Instruct [ 106 ] , evaluated under identical settings with five independent runs.
Setting
LD50
hERG C
Carc L
Tox21
AMES
ToxCast
ClinTox
DILI
hERG K
hERG
SkinRxn
Overall
Multimodal
0.3 ± 0.7
61.0 ± 4.7
34.0 ± 1.5
45.0 ± 1.8
20.0 ± 1.1
52.3 ± 4.2
47.7 ± 1.4
5.3 ± 0.7
33.3 ± 4.6
14.3 ± 0.9
21.7 ± 1.3
30.5 ± 0.7
Text-only
0.3 ± 0.7
56.3 ± 4.8
34.3 ± 3.0
45.0 ± 3.0
16.3 ± 0.7
51.0 ± 1.7
44.3 ± 2.0
4.7 ± 0.7
31.0 ± 3.4
12.3 ± 1.8
22.0 ± 2.5
28.9 ± 0.7
Table 5: Task-level success rates (%) of multimodal and text-only models at test time. We compare GPT-4.1 [ 108 ] under the standard multimodal setting (SMILES + 2D molecular image) against a text-only variant (SMILES only). Reported values are mean ± std over five independent runs, evaluated under identical prompts, decoding settings, and the ToxiEval pipeline.
Appendix figures & tables14 assets
Supplementary material from the paper’s appendix.
Appendix
Task
Toxicity Mechanism
AMES
Mutagenicity measured by Ames bacterial reverse-mutation assay (DNA damage/frameshift mutations).
hERG
Cardiotoxicity liability via blockade of the hERG potassium channel (QT prolongation/arrhythmia risk).
hERG_Karim
hERG channel blockade liability (binary; thresholded around <10μ M vs. ≥10μ M).
hERG_Central
hERG channel inhibition profiling (percent inhibition/blocker classification from hERGCentral assays).
Tox21
Target/pathway-specific in vitro toxicity across 12 assays (nuclear receptor signaling and stress-response pathways).
ToxCast
High-throughput toxicity screening across >600 in vitro assays spanning diverse biological targets/pathways.
Appendix
Table 6: Toxicity mechanism descriptions of the main tasks.
Figure 7: Word cloud visualization of the nearly 30 toxicity mechanisms covered by the main tasks and subtasks in ToxiMol.
Task ID
Subtask Name
Toxicity Mechanism
T1
Tox21_SR_ARE
Oxidative stress response via Antioxidant Response Element (ARE)
T2
Tox21_SR_p53
Genotoxic stress-induced activation of p53 tumor suppressor pathway
T3
Tox21_NR_Aromatase
Endocrine disruption via inhibition of aromatase enzyme
T4
Tox21_SR_ATAD5
DNA replication stress signaling through ATAD5 gene expression
Table 8: List of ToxCast Subtasks and Corresponding Toxicity Mechanisms.
Prompt Template
hERG
hERG C
hERG K
Original
26.7
78.3
51.7
Unified
26.7
81.7
53.3
Appendix
Table 9: Comparison of success rates (%) on hERG-related tasks using the original versus unified prompt templates. Results are obtained with Claude-3.7 Sonnet [ 2 ] as the repair generator, evaluated under the ToxiEval pipeline with TxGemma-Predict [ 41 ] . The “Unified” setting applies a single shared prompt template (derived from the hERG C format) to all three tasks while keeping the test set and decoding configuration unchanged.
Figure 8: Overall prompt design for detoxification tasks using MLLMs.
Figure 9: Task prompt design for detoxification tasks of Tox21 using MLLMs.
Symbol
Definition
T
Toxicity repair task identifier (e.g., AMES, DILI, hERG).
Sraw={SMILESiraw}i=1n
Set of toxicity-positive SMILES strings requiring repair.
IMGi
2D molecular structure image associated with SMILESiraw , rendered via RDKit.
Pbase
Base prompt template encoding MLLM role, objectives, and formatting rules.
Ptask
Task-level annotation describing toxicity mechanism and optimization focus.
Psubtask
Subtask-level annotation for fine-grained tasks (e.g., specific Tox21 endpoints).
Appendix
Table 11: Symbol Definitions for Mechanism-Aware Prompt Annotation Pipeline.
Symbol
Definition
SMILESiraw
Original toxicity-positive molecule requiring repair.
Candidate repaired molecules generated by the MLLM for input i .
s∈Sirep
A single candidate molecule from the repaired set.
T
Toxicity repair task identifier (e.g., AMES, DILI, hERG, LD50).
Gtox
Toxicity prediction model used for safety scoring (e.g., TxGemma-Predict).
Ssafe
Safety score of s based on task T (binary or normalized).
Appendix
Table 12: Symbol Definitions for the ToxiEval Evaluation Framework.
Models
T1
T2
T3
T4
T5
T6
T7
T8
T9
T10
T11
T12
Domain-Specific MLLMs
ChemVLM 26B [ 67 ]
0.0
0.0
40.0
0.0
80.0
20.0
100.0
0.0
80.0
0.0
0.0
60.0
ChemVLM 8B [ 67 ]
0.0
0.0
40.0
0.0
100.0
40.0
100.0
0.0
60.0
20.0
20.0
80.0
Intern-S1 [ 98 ]
0.0
20.0
40.0
40.0
100.0
20.0
100.0
20.0
80.0
20.0
60.0
80.0
Intern-S1 Mini [ 98 ]
0.0
0.0
60.0
20.0
100.0
40.0
100.0
20.0
80.0
40.0
40.0
40.0
Average Success Rate
0.0
5.0
45.0
15.0
95.0
40.0
100.0
10.0
75.0
30.0
25.0
65.0
Appendix
Table 13: Success Rates of Each Model on the 12 Sub-Tasks of Tox21.
Models
T1
T2
T3
T4
T5
T6
T7
T8
T9
T10
Domain-Specific MLLMs
ChemVLM 26B [ 67 ]
66.7
33.3
50.0
16.7
0.0
16.7
33.3
16.7
16.7
33.3
ChemVLM 8B [ 67 ]
83.3
50.0
50.0
50.0
16.7
16.7
33.3
33.3
66.7
33.3
Intern-S1 [ 98 ]
83.3
66.7
66.7
50.0
16.7
16.7
66.7
66.7
66.7
50.0
Intern-S1 Mini [ 98 ]
66.7
83.3
66.7
50.0
16.7
16.7
66.7
66.7
66.7
50.0
Average Success Rate
75.0
58.3
54.2
41.7
12.5
16.7
41.7
37.5
58.3
41.7
Appendix
Table 14: Success Rates of Each Model on the 10 Selected Sub-Tasks of ToxCast.
Models
LD50
hERG C
Carc L
Tox21
AMES
ToxCast
ClinTox
DILI
hERG K
hERG
SkinRxn
Overall
Domain-Specific MLLMs
ChemVLM 26B [ 67 ]
98.3
55.0
31.7
68.3
11.7
66.7
71.7
18.3
25.0
20.0
20.0
44.2
ChemVLM 8B [ 67 ]
100.0
45.0
38.3
71.7
25.0
83.3
66.7
11.7
21.7
18.3
21.7
45.8
Intern-S1 [ 98 ]
100.0
80.0
66.7
80.0
43.3
85.0
80.0
11.7
51.7
26.7
43.3
60.8
Intern-S1 Mini [ 98 ]
95.0
78.3
65.0
71.7
38.3
83.3
81.7
11.7
38.3
31.7
36.7
57.4
Average Success Rate
98.3
64.6
50.4
72.9
29.6
79.6
75.0
13.3
34.2
24.2
30.4
52.0
Appendix
Table 15: Toxicity Repair Valid Rates (%). hERG C , hERG K , and Carc L refer to the hERG_Central , hERG_Karim , and Carcinogens tasks, respectively. SkinRxn denotes the SkinReaction task.
Figure 10: Representative successful toxicity repair cases of Claude Opus 4.5 [ 110 ] .
Figure 11: Representative failed toxicity repair cases of Claude Opus 4.5 [ 110 ] .
Molecular optimization computationally designs or edits chemical structures to improve desired properties while respecting chemical and practical constraints.
Large Language Models (LLMs) and Vision Language Models (VLMs) have recently shown promising capabilities in various scientific domain. In particular, these advances have opened new opportunities in drug discovery, where the ability to understand and modify molecular structures is critical for optimizing drug properties such as efficacy and toxicity. However, existing models and benchmarks often overlook toxicity-related challenges, focusing primarily on general property optimization without adequately addressing safety concerns. In addition, even existing toxicity repair benchmarks suffer from limited data diversity, low structural validity of generated molecules, and heavy reliance on proxy models for toxicity assessment. To address these limitations, we propose MolDeTox, a novel benchmark for molecular detoxification, designed to enable fine-grained and reliable evaluation of toxicity-aware molecular optimization across stepwise tasks. We evaluate a wide range of general-purpose LLMs and VLMs under diverse settings, and demonstrate that understanding and generating molecules at the fragment-level improves structural validity and enhances the quality of generated molecules. Moreover, through detailed task-level performance analysis, MolDeTox provides an interpretable benchmark that enables a deeper understanding of the detoxification process. Our dataset is available at : https://huggingface.co/datasets/MolDeTox/MolDeTox
Jueon Park, Wonjune Jang, Jiwoo Lee +2
Korea University · Myongji University · AIGEN Sciences
Recent advances in large language models (LLMs) have enabled molecular reasoning for property prediction. However, toxicity arises from complex biological mechanisms beyond chemical structure, necessitating mechanistic reasoning for reliable prediction. Despite its importance, current benchmarks fail to systematically evaluate this capability. LLMs can generate fluent but biologically unfaithful explanations, making it difficult to assess whether predicted toxicities are grounded invalid mechanisms. To bridge this gap, we introduce ToxReason, a benchmark grounded in the Adverse Outcome Pathway (AOP) that evaluates organ-level toxicity reasoning across multiple organs. ToxReason integrates experimental drug-target interaction evidence with toxicity labels, requiring models to infer both toxic outcomes and their underlying mechanisms from Molecular Initiating Event (MIE) to Adverse Outcome (AO). Using ToxReason, we evaluate toxicity prediction performance and reasoning quality across diverse LLMs. We find that strong predictive performance does not necessarily imply reliable reasoning. Furthermore, we show that reasoning-aware training improves mechanistic reasoning and, consequently, toxicity prediction performance. Together, these results underscore the necessity of integrating reasoning into both evaluation and training for trustworthy toxicity modeling.
Jueon Park, Wonjune Jang, Chanhwi Kim +2
Korea University · Myongji University · University of Texas Health Science Center at Houston +1
Molecular toxicity prediction is increasingly used to prioritise compounds before experimental testing, but conventional benchmark performance can overstate practical utility when structurally related molecules occur across training and test folds. We introduce ToxLens, a reproducible multi-task graph-learning framework for 11 toxicity endpoints spanning Ames mutagenicity, acute oral toxicity, hERG inhibition, and Tox21 nuclear-receptor and stress-response assays. The workflow combines conservative chemical curation, sphere-exclusion filtering, a leakage-aware UMAP-HDBSCAN split, parallel graph and global-feature encoders joined by late concatenation, temperature-scaled Monte Carlo dropout with conformal-style prediction sets, applicability-domain analysis, and SHAP-guided toxicophore discovery with occlusion controls. On the leakage-controlled test fold, a five-seed soft-voting ensemble achieved a Matthews correlation coefficient score of 0.44, an area under the receiver operating characteristic curve score of 0.83, and an area under the precision-recall curve score of 0.58. It exceeded four ECFP4-based shallow baselines on all 11 endpoints under the same split and validation-based threshold-selection protocol. Controlled ablations showed that the global pathway was important, whereas late concatenation outperformed the tested gated and feature-wise linear modulation fusion variants. Conformal-style prediction sets revealed substantial endpoint-specific variation in set efficiency, and discrimination and calibration improved with similarity to the training domain. Retraining on fixed published Tox21 Challenge and TDA folds produced competitive, but not uniformly state-of-the-art, performance. SHAP-guided occlusion and consensus subgraph mining yielded model-derived structural hypotheses, 44 of which contained at least one occurrence that passed the predefined counterfactual criteria.
Magnus H. Strømme, Alex G. C. de Sá, David B. Ascher
The Australian Centre for Ecogenomics, School of Chemistry and Molecular Biosciences, The University of Queensland, Brisbane, Queensland, 4072, Australia · Computational Biology and Clinical Informatics, Baker Heart and Diabetes Institute, Melbourne, Victoria, 3004, Australia