ASCCT Award Winners Webinar
Registration Link: https://attendee.gotowebinar.com/register/9200492703200567307
Two presentations will be offered on March 11th, 2020 by the winners of the ASCCT award.
Wednesday, March 11th, 2020
11:00 AM US ET
Toxicological mechanistic inference: Generating mechanistic explanations of adverse outcomes
Presenter: Ignacio J. Tripodi, PhD Candidate in Computer Science / Interdisciplinary Quantitative Biology, University of Colorado, Boulder
Government regulators and others concerned about toxic chemicals in the environment hold that a mechanistic, causal explanation of a mechanism of toxicity is strongly preferred over a statistical or machine learning-based prediction by itself. We thus present a mechanistic inference framework, which can generate hypotheses of the most likely mechanisms of toxicity for a specific chemical and cell type, using gene expression time series on human tissue and a semantically-interconnected knowledge graph. We seek enrichment in our manually-curated list of high-level mechanisms of toxicity (e.g. "Triggering of caspase-mediated apoptosis via release of cytochrome C", or "Mitochondria-mediated toxicity by inhibition of electron transport chain"), represented as causally-linked Gene Ontology concepts.
Our knowledge representation is an extension of the PheKnowLator knowledge graph. It consists of an integration of concepts from multiple ontologies (GO, PRO, HPO, ChEBI, PATO, DOID, CL), as well as relevant concepts from Reactome, Uniprot, the cellular toxicogenomics database (CTD), and the AOP Wiki. The expression assays were obtained from the Open TG-Gates, CarcinoGenomics, and other projects, consisting of human liver, kidney, and lung, bronchial, buccal, and nasal epithelial cells exposed to a sizeable number of chemicals that elicit different mechanisms of toxicity. Both our knowledge graph and experimental transcriptomics data are human-centric. Besides predicting the most likely mechanisms of toxicity from the transcriptomics assays, we generate putative explanations based on the most significant genes at each time point with known links to their corresponding mechanism steps. This provides a transparent, possible explanation for the mechanisms of toxicity, that would help a researcher’s decision-making process and aid further experimental design. Furthermore, we were able to experimentally validate our mechanistic predictions for some chemicals without an established mechanism of toxicity.
Co-expression Network Analysis to Identify Heterogeneity Between the Breast Cancer Cell Line MCF-7 and Human Breast Cancer Tissues
Presenter: Vy Tran, PhD Student, Center for Alternatives to Animal Testing, Johns Hopkins University, USA
Lack of characterization of cell lines can have serious consequences on the translatability of in vitro scientific studies to human applications. This project focuses on the Michigan Cancer Foundation-7 (MCF-7) cells, a human breast adenocarcinoma cell line that is commonly used for in vitro cancer research, with over 33,000 publications in PubMed. Previously, our lab has shown that even MCF-7 cells obtained from the same cell batch at the same cell bank can display cellular and phenotypic heterogeneity, which affected reproducibility of experiments using this cell line. Using publicly available data sets, this study explored the key similarities and differences in gene expression networks of MCF-7 cell lines and human breast cancer tissues using Weighted Gene Correlation Network Analysis (WGCNA) – a method that takes advantages of correlation amongst genes and graph theory – and functional annotation. Our results indicate that results from MCF7 have to be used extremely cautiously as a proxy for human breast cancer physiology. To begin with, there is minimal overlap of gene expression levels even at the most basic metric of similarity (sorting genes by expression level). Moreover, when using a more sophisticated metric, such as weighted gene correlation network analysis, the hub genes are substantially different and more importantly, there are several genes that are "druggable" that are apparent in the BRCA dataset that would've been missed if relying solely on MCF7. In conclusion, our data cautions against the advisability of scaling-up transcriptomic data - in the case of MCF7 cells, owing to the genetic instability as well as artifacts intrinsic in tissue culture and transcriptomic data - a "big-data" approach may simply adding more noise rather than signal. Therefore, MCF7 cells -like all models - should be used in tandem with parallel approaches to provide added confidence in the data.