Screening agrochemicals and pharmaceuticals for liver toxicity is expensive and time-consuming. This study uses comparative supervised machine learning on the rat liver TG-GATEs dataset to identify early exposure gene signatures and predictive models. Ten gene biomarkers were identified using three feature selection methods, predicting liver necrosis with high accuracy in an independent validation dataset (MAQC-II). Nine of these genes are involved in metabolism and detoxification, and one in transcriptional regulation. Several are also implicated in liver carcinogenesis. This biomarker gene signature offers high accuracy and could accelerate toxicity testing.

Brandi Patrice Smith, Loretta Sue Auvil, Michael Welge, Colleen Bannon Bushell, Rohit Bhargava, Navin Elango, Kamin Johnson, Zeynep Madak-Erdogan