Impact of gene selection criteria on transcriptomic ontology-based point of departure estimates

Abstract

Apical effects are typically associated with changes in gene expression, which allows for the use of short- term in life transcriptomic studies to derive biologically relevant points of departure (POD). These methods offer cost savings over conventional toxicology assessments and can derive data from very short-term studies where apical effects may not yet be present. When there is limited or insufficient data for a conventional POD assessment, a transcriptomic screen could provide valuable data for deriving a cellular bioactivity POD for chemical screening and hazard assessment. We used existing transcriptomic data from published 5-day rat in vivo kidney and liver exposures to examine the effect of differential gene expression metrics for the selection of genes used for ontology pathway-based POD derivation. Williams Trend Test (WTT) indicate no gene expression dose-response in 6 instances and ANOVA in one, while DESeq2 detected differentially expressed genes in all instances. The three statistical metrics produced consistent POD values. One chemical (PFOA in liver) showed ontology enrichment indicative of a cytotoxic response at the highest dose, emphasizing the effect which too high a dose can have on the derivation of POD values if such response is not accounted for. Whether the choice of a gene selection metric combining both a statistical significance criterion as well as a minimum magnitude of change threshold affects the sensitivity of POD values depends on the specifics of the dose- response. Existing alternative and complementary analyses could be utilized with existing analyses pipelines to better inform analytical decisions when using transcriptomics and BMD for point of departure determinations.