期刊
ENVIRONMENTAL SCIENCE & TECHNOLOGY
卷 57, 期 33, 页码 12291-12301出版社
AMER CHEMICAL SOC
DOI: 10.1021/acs.est.3c02792
关键词
supervised learning; bigdata; data mining; adverse outcome pathway modeling; biological pathwaymodeling
Animal models are not effective in predicting hepatotoxicity in humans, leading to the development of biological pathway-based alternatives such as in vitro assays. Public screening programs have tested thousands of chemicals using high-throughput screening assays. Developing pathway-based models for complex toxicities like hepatotoxicity remains challenging. This study aimed to develop a computational strategy for developing pathway-based models for complex toxicities.
Failure of animalmodels to predict hepatotoxicity in humans hascreated a push to develop biological pathway-based alternatives, suchas those that use in vitro assays. Public screening programs (e.g.,ToxCast/Tox21 programs) have tested thousands of chemicals using invitro high-throughput screening (HTS) assays. Developing pathway-basedmodels for simple biological pathways, such as endocrine disruption,has proven successful, but development remains a challenge for complextoxicities like hepatotoxicity, due to the many biological eventsinvolved. To this goal, we aimed to develop a computational strategyfor developing pathway-based models for complex toxicities. Usinga database of 2171 chemicals with human hepatotoxicity classifications,we identified 157 out of 1600+ ToxCast/Tox21 HTS assays to be associatedwith human hepatotoxicity. Then, a computational framework was usedto group these assays by biological target or mechanisms into 52 keyevent (KE) models of hepatotoxicity. KE model output is a KE scoresummarizing chemical potency against a hepatotoxicity-relevant biologicaltarget or mechanism. Grouping hepatotoxic chemicals based on the chemicalstructure revealed chemical classes with high KE scores plausiblyinforming their hepatotoxicity mechanisms. Using KE scores and supervisedlearning to predict in vivo hepatotoxicity, including toxicokineticinformation, improved the predictive performance. This new approachcan be a universal computational toxicology strategy for various chemicaltoxicity evaluations. Acomputational framework capable of modeling biologicalpathways that integrates with toxicokinetic information is described.This modeling framework can identify potential hepatotoxicants andtheir mechanisms of hepatotoxicity.
作者
我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。
推荐
暂无数据