{"doi":"10.1016/j.patter.2022.100586","title":"Predicting drug toxicity at the intersection of informatics and biology: DTox builds a foundation","abstract":"Hao et al. (2022) present DTox (deep learning for toxicology), a neural network designed to predict and probe the sites and potential mechanisms underlying chemical toxicity; results provide a map to facilitate modular testing and improvements across multiple disparate applications. Hao et al. (2022) present DTox (deep learning for toxicology), a neural network designed to predict and probe the sites and potential mechanisms underlying chemical toxicity; results provide a map to facilitate modular testing and improvements across multiple disparate applications. Main textMost, if not all, readers of Patterns would likely agree that progress in medicine in the coming years will increasingly depend on improved use of data resources. Historically, each advance in both data acquisition and analysis has improved our insight into disease and our ability to predict, diagnose, and treat it. This historical trend ranges from \"simple\" advances in clinical chemistry, such as the ability to measure glucose (diabetes), to modern methods that identify genetic abnormalities predisposing individuals to diseases (e.g., BRCA1 mutations and breast cancer1Futreal P.A. Liu Q. Shattuck-Eidens D. Cochran C. Harshman K. Tavtigian S. Bennett L.M. Haugen-Strano A. Swensen J. Miki Y. et al.BRCA1 mutations in primary breast and ovarian carcinomas.Science. 1994; 266: 120-122https://doi.org/10.1126/science.7939630Crossref PubMed Scopus (1130) Google Scholar). Critically, this historical trend is equally apparent in data analysis—witness examples such as the role of statistics in seeking to establish the role of chance versus signal (e.g., clinical trials), information theory, signal processing (establishing the role of randomness and the limitations imposed by noise on detectable signal), the progressive improvements in regression-based approaches (i.e., from linear regression to least angle regression), and other modeling approaches (e.g., projection methods, trees, ensembles, SVMs, and neural-networks, including their use in deep learning and AI). Modern biomedical research would be crippled without these breakthroughs. Thus, substantial historical precedent suggests that new informatics technologies will open new insights and approaches into human health and disease. In this issue of Patterns, Hao et al.2Hao Y. Romano J.D. Moore J.H. Knowledge-guided deep learning models of drug toxicity improve interpretation.Patterns. 2022; 3: 100565Abstract Full Text Full Text PDF Google Scholar provide just such an advance to help predict not only which potential drug candidates will have side effects but where such toxicity will manifest at the level of the individual, the organ system, and the cell.Improving toxicity predictions has important implications for society (e.g., costs, delayed development, abandoned programs), for the individual (e.g., trial subjects), and for the technology itself, as better understanding can often be directly parlayed into improving drug candidates/pharmacophores, co-development, or other regimens. In silico drug screening has had successes in both increasing primary hit rates and predicting some drugs’ toxicity, but the latter are often black-box models that provide little added actionable (e.g., mechanistic pathways) information. Approaches designed to identify the elements of models that contribute to class discernment (e.g., LIME,3Ribeiro M.T. Singh S. Guestrin C. Why should i trust you? Explaining the predictions of any classifier.Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining (KDD '16). Association for Computing Machinery, 2016: 1135-1144https://doi.org/10.1145/2939672.2939778Crossref Scopus (4870) Google Scholar Shapley4Lundberg S.M. Lee S.-I. A unified approach to interpreting model predictions.in: Proceedings of Advances in Neural Information Processing Systems. 30. 2017: 4765-4774Google Scholar) are powerful for recognizing the mathematical drivers behind such classificatio","journal":"Patterns","year":2022,"id":311278,"datarank":0.0,"base_score":0.0,"endowment":0.0,"self_citation_contribution":0.0,"citation_network_contribution":0.0,"self_endowment_contribution":0.0,"citer_contribution":0.0,"corpus_percentile":null,"corpus_rank":null,"citation_count":0,"citer_count":0,"citers_with_citation_signal":0,"citers_with_endowment":0,"datacite_reuse_total":0,"is_dataset":false,"is_dataset_confidence":0.9348,"is_data_producer":false,"deposit_databanks":null,"is_oa":true,"file_count":0,"downloads":0,"has_version_chain":false,"published_date":"2022-01-01","fair_score":null,"fair_percentile":null,"algorithm_id":"datarank_citation_only_1hop_v6","ranking_scope":"data_only","authors":[{"id":75182,"name":"Bruce S. Kristal","orcid":"0000-0001-6103-7745","position":1,"is_corresponding":false},{"id":783794,"name":"Matthew J. Sniatynski","orcid":"0000-0003-2915-6677","position":0,"is_corresponding":true}],"reference_count":12,"raw_metadata":{"citation_network_status":"fetched"},"created_at":"2026-07-19T00:33:28.480200Z","pmid":"36124303","pmcid":null,"fwci":null,"citation_percentile":null,"influential_citations":0,"oa_status":null,"license":null,"views":0,"total_file_size_bytes":0,"version_count":0,"fair_f":null,"fair_a":null,"fair_i":null,"fair_r":null,"fair_zscore":null,"fair_rationale":null,"fair_model":null,"fair_agent_version":null,"fair_fulltext_source":null,"fair_has_llm":null,"fair_computed_at":null,"clinical_trials":[],"software_tools":[],"db_accessions":[],"linked_datasets":[],"topics":[]}