Although there have been nationwide reductions in recent cancer death, liver cancer mortality remains to be a problem in rural Illinois. Researchers anticipate that exposure to environmental toxicants in rural Illinois, may be possible causes of increased liver cancer mortality in females. The purpose of our study is to identify potential biomarkers, which are related to early liver toxicity through machine learning. We accessed gene expression data from the open source Toxicogenomics Project-Genomics Assisted Toxicity Evaluation System (TG-GATES) database from rats that had been exposed to over 170 toxicants relevant to liver cancer. We performed a differential gene expression analysis to identify significant features or genes, which are related to hepatotoxicity endpoints, specifically necrosis. Then we built our predictive model using supervised machine learning (ML) through feature selection and classification models. We performed differential gene expression analysis and feature selection to decrease the dimensionality of the gene set. We then used classification modeling to classify genes related to necrosis. Lastly, we tested our predictive model, on an independent data set. Our model predicted specific genes that were highly related to liver toxicant exposure.
Note: This abstract was not presented at the meeting.
Citation Format: Brandi Smith, Zeynep Madak-Erdogan. A machine learning based approach to identify biomarkers of environmental toxicant exposures relevant to liver cancer disparities in rural Illinois [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2019; 2019 Mar 29-Apr 3; Atlanta, GA. Philadelphia (PA): AACR; Cancer Res 2019;79(13 Suppl):Abstract nr 5124.