ToxLens is introduced, 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 and reveals substantial endpoint-specific variation in set efficiency and discrimination and calibration improved with similarity to the training domain.
Abstract
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.
Despite breakthroughs in predicting acute organ toxicities, advanced computational models continue to struggle with complex, insidious end points like drug-induced bone toxicity. A fundamental limitation of most supervised pipelines is their failure to explicitly model intermolecular similarities during task-specific t...
He Huang, Qin-Yi Wang, Man-Zhan Zhang et al.· JACS Au· 0 citations
A unified and standardized benchmarking framework for GDL-based toxicity prediction is introduced, systematically evaluating more than 20 representative approaches under consistent experimental conditions and across multiple datasets and partitioning strategies, enabling a fair and reproducible comparison of model perf...
Noel Suarez-Barro, M. Lama, J. C. Vidal· 0 citations
A causality-integrated graph learning framework that embeds toxicological mechanisms extracted from large-scale literature mining as directed graphs within deep learning (DL) models, establishing a mechanism-constrained, causality-informed learning paradigm that is highly relevant for multi-endpoint toxicity assessment...
Hao-Yue Tan, Tong Bao, Yin Fang et al.· Proceedings of the National...· 0 citations
To address widespread data leakage and inconsistent evaluation in protein–ligand affinity prediction, we introduce PLABench, a leakage-controlled and target-centric benchmarking framework that enables standardized comparison across sequence- and structure-based models. We benchmark nine deep learning methods across bli...
Accurately predicting the acute toxicity of chemicals to Daphnia magna is critical for environmental risk assessment. Yet, current deep learning models often exhibit limited generalization and overconfidence when processing out-of-distribution molecules, risking dangerous false-negative errors due to scarce training da...
Ze-Xin Wen, Rong Yang, Rui-Dong Chen et al.· Environmental Pollution· 0 citations
Molecular property prediction is a key task in AI-driven drug discovery, yet the prevalence and impact of label imbalances in molecular property regression remain poorly understood. Through a systematic benchmark of widely used molecular property data sets, we show that target values are often highly imbalanced and tha...
Y. Sun, Yu Shi, Alana Deng et al.· Journal of Chemical Informat...· 0 citations
Related blog posts
MIT News · Artificial Intelligence· news.mit.eduNov 17, 2026
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