HExpPredict: In Vivo Exposure Prediction of Human Blood Exposome Using a Random Forest Model and Its Application in Chemical Risk Prioritization

BACKGROUND: Due to many substances in the human exposome, there is a dearth of exposure and toxicity information available to assess potential health risks. Quantification of all trace organics in the biological fluids seems impossible and costly, regardless of the high individual exposure vari-ability. We hypothesized that the blood concentration (CB) of organic pollutants could be predicted via their exposure and chemical properties. Developing a prediction model on the annotation of chemicals in human blood can provide new insight into the distribution and extent of exposures to a wide range of chemicals in humans.OBJECTIVES: Our objective was to develop a machine learning (ML) model to predict blood concentrations (CBs) of chemicals and prioritize chemi-cals of health concern.METHODS: We curated the CBs of compounds mostly measured at population levels and developed an ML model for chemical CB predictions by con-sidering chemical daily exposure (DE) and exposure pathway indicators (dij), half-lives (t1=2), and volume of distribution (Vd). Three ML models, including random forest (RF), artificial neural network (ANN) and support vector regression (SVR) were compared. The toxicity potential or prioriti-zation of each chemical was represented as a bioanalytical equivalency (BEQ) and its percentage (BEQ%) estimated based on the predicted CB and ToxCast bioactivity data. We also retrieved the top 25 most active chemicals in each assay to further observe changes in the BEQ% after the exclusion of the drugs and endogenous substances. RESULTS: We curated the CBs of 216 compounds primarily measured at population levels. RF outperformed the ANN and SVF models with the root mean square error (RMSE) of 1.66 and 2.07 lM, the mean absolute error (MAE) values of 1.28 and 1.56 lM, the mean absolute percentage error (MAPE) of 0.29 and 0.23, and R2 of 0.80 and 0.72 across test and testing sets. Subsequently, the human CBs of 7,858 ToxCast chemicals were suc-cessfully predicted, ranging from 1.29 x 10-6 to 1.79 x 10-2 lM. The predicted CBs were then combined with ToxCast in vitro bioassays to prioritize the ToxCast chemicals across 12 in vitro assays with important toxicological end points. It is interesting that we found the most active compounds to be food additives and pesticides rather than widely monitored environmental pollutants.DISCUSSION: We have shown that the accurate prediction of internal exposure from external exposure is possible, and this result can be quite use-ful in the risk prioritization. https://doi.org/10.1289/EHP11305

Automated summary

A machine learning model was developed to predict the concentration of chemicals in blood and assess their potential health impacts. The study found that food additives and pesticides were the most active compounds, rather than widely monitored environmental pollutants.