Integration of Transcriptomic Point of Departure Metrics Into the MoAviz Visualization Framework

Download Poster PDF: MoAviz SOT 2020 Poster PDF

Authors: M. B. Black1, K. Bronson 1, S. N. Pendse 1, J. Fitzpatrick 1, R. A. Clewell2, M. E. Andersen 1, P. D. McMullen 1

Affiliations: 1 ScitoVation LLC, Durham, NC 27713 & 2 21st Century Tox Consulting, Chapel Hill, NC, USA

Gene expression profiling is emerging as a viable way to evaluate mode of action and points of departure and has the potential to drastically reduce testing costs and product development time. The first steps of using transcriptional responses as a basis of safety assessment are being taken. However, several important considerations remain unresolved, including how to translate expression changes into adverse outcome pathways or other definitions of mode of action, and the best manner with which to summarize gene expression data into a point of departure. Recently we developed an interactive browser application, MoAviz, to facilitate the examination of gene expression data across dose and time for mode of action studies of chemical perturbation for 204 compounds (spanning 290 million gene expression change values). We used MoAviz to quantitatively compare pathway-level transcriptomic signatures across compounds with well-known modes of action, and across different model systems, providing the groundwork for performing “biological read-across” between compounds based on their transcriptomic fingerprints. We evaluated the extent to which gene expression changes from in-life exposures could be associated with mode of action by developing a novel similarity index—the Modified Jaccard Index (MJI)—that provides a quantitative description of genomic pathway similarity. While typical compound-compound similarity is low (MJI = 0.026), clustering of the TG-GATES compounds identifies groups of similar compounds. Some clusters aggregated compounds with known similar modes of action, including PPARα agonists (MJI = 0.330) and NSAIDs (MJI = 0.327). We continue to extend the MoAviz interface and database by incorporating whole transcriptome benchmark dose analyses and point of departure (POD) summary, including the command line modeling features of the BMDExpress2 software. This integration will include statistical pre-filtering of transcriptomic gene expression data, dose response modeling of individual genes, ontology over-representation of genes, and POD summary based on current proposed best practices for gene-based and pathway-based derivation of POD. By combining mode of action and POD tools in an interactive interface, MoAviz will facilitate the use of transcriptomics data over a variety of chemical safety contexts.