Evaluating human health impacts of emerging environmental contaminants using artificial intelligence

The growing number of new environmental pollutants is an immense burden on human health risk assessment where rate of introducing chemicals and its detection in the environment is out-running the traditional toxicological tools of assessment. Artificial intelligence methods of data collection provide an opportunity to fill this gap by combining heterogeneous information that applies to hazard and exposure. In paper, a designed analytical framework of artificial intelligence was created to evaluate and rank the possible human health hazards of the representative emerging contaminants of the chosen artificial intelligence through combined physicochemical descriptors, toxicological bioassay indicators, and exposure related parameters. Several simple machine learning models such as logistic regression, random forest, gradient boosting and deep neural networks were trained and assessed through repeated cross validation to offer statistical soundness and to reduce overfitting. Accuracy, precision, recall, F1 score and area under the receiver operating characteristic curve were used to measure model performance. The ensemble random forest model had shown excellent and statistically significant performance with an average classification accuracy of about 85 percent, recall of more than 0.85 with high-risk contaminants and an area under curve of near 0.90 with each fold of validation showing strong discriminating capacity. Explainable artificial intelligence analyses indicated that least amount of lipophilicity, environmental persistence, signs of endocrine related biological activity, and the intensity of use or production were the most important contributors to risk classification and together, they explained most of the model explaining power. Comprehensively, these results can be used to understand that statistically sound artificial intelligence models can be successful in the recognition and rank of potential human health interest emerging contaminants.