抄録
An increase in liver weight is a relatively common finding experienced in preclinical repeat dose studies of new drug candidates. Occasionally, liver weight increase accompanies characteristic histopathological changes that are typified by ground-glass appearance or eosinophilic granular degeneration of hepatocytes. Generally, the onset of these changes is revealed in repeat-dose studies by histopatholoical evaluations. Thus, we aimed to predict and distinguish between the different types of liver weight increase using toxicogenomic profiling in single-dose studies.
We obtained large-scale transcriptome data of the rat liver in single-dose studies of 150 compounds from Genomics Assisted Toxicity Evaluation system (TG-GATEs) developed by Toxicogenomics Project in Japan. Based on the histopathological findings of the liver in repeat-dose studies, gene expression profiles were divided into 3 classes. Filtertype gene selection was applied to 3-class training samples, and a linear classifier was constructed using the selected genes. The accuracy of the classifier was evaluated by 5-fold cross-validation. Our validation showed a sensitivity of 80.3% with 9.7% false positives for the discriminant model of ground-glass appearance and a sensitivity of 100% with no false positives for the degeneration of eosinophilic granules model.
Toxicogenomics-based discrimination in short-term studies has the potential to assist in toxicological evaluation of liver weight increase.
