Machine learning on drug-specific data to predict small molecule teratogenicity

Abstract

1Pregnant women are an especially vulnerable population, given the sensitivity of a developing 2 fetus to chemical exposures. However, prescribing behavior for the gravid patient is guided on 3 limited human data and conflicting cases of adverse outcomes due to the exclusion of pregnant 4 populations from randomized, controlled trials. These factors increase risk for adverse drug 5 outcomes and reduce quality of care for pregnant populations. Herein, we propose the 6 application of artificial intelligence to systematically predict the teratogenicity of a prescriptible 7 small molecule from information inherent to the drug. Using unsupervised and supervised 8 machine learning, our model probes all small molecules with known structure and teratogenicity 9 data published in research-amenable formats to identify patterns among structural, meta-10 structural, and in vitro bioactivity data for each drug and its teratogenicity score. With this 11 workflow, we discovered three chemical functionalities that predispose a drug towards increased 12 teratogenicity and two moieties with potentially protective effects. Our models predict three 13 clinically-relevant classes of teratogenicity with AUC = 0.8 and nearly double the predictive 14 accuracy of a blind control for the same task, suggesting successful modeling. We also present 15 extensive barriers to translational research that restrict data-driven studies in pregnancy and 16 therapeutically "orphan" pregnant populations. Collectively, this work represents a first-in-kind 17 platform for the application of computing to study and predict teratogenicity. 18