Dissertation/Thesis Abstract

Inferring Mechanisms of Toxicity from Differential Genomics and Semantic Knowledge Representations
by Tripodi, I. J., Ph.D., University of Colorado at Boulder, 2020, 149; 27996986
Abstract (Summary)

This thesis explores a combination of genomics analysis and semantic knowledge representation useful for computational toxicology. I first discuss a novel approach to infer differences in transcription factor (TF) activity between biological conditions, utilizing a variety of omics assays. This method for genome-wide exploratory analysis in exposure studies is valid for a simple and inexpensive protocol (ATAC-seq). It can also be used to study transcriptional perturbations by toxicants that may target protein receptors, at very early time points and with a fine time resolution, via more sophisticated protocols (GRO-seq, PRO-seq). Detecting changes between biological conditions is only half the problem. By integrating public databases, I provide ways to relate the highlighted TFs in different dimensions, thus expediting downstream analysis. Further exploiting the power of ATAC-seq, I show that there are inherent signatures in the peaks from ATAC-seq signal that, combined with the underlying sequence, can be used to predict the presence of nascent transcription or histone modifications at that genomic coordinate.

In addition to genomics analysis, I explore applications of semantic knowledge representation. I demonstrate how a consistent integration of data from public databases and open biomedical ontologies can be used to infer novel drug-drug interactions, chemical-protein relations, or enrichment of mechanisms of toxicity. A significant portion of computational toxicology work focuses on the prediction of outcomes, rather than the generation of mechanistic explanations for said outcomes (thus sometimes being perceived as a "black box"). I show it's possible to produce putative explanations for our predictions of cellular toxicity modes of action from experimental data. The mechanism enrichment strategy accounts for the sequential order in which measured biological events happen. Here, the measured phenomena are changes in gene expression, however this mechanistic inference framework can be adapted to other types of mechanisms beyond toxicology.

Indexing (document details)
Advisor: Dowell, Robin D., Hunter, Lawrence E.
Commitee: Larremore, Daniel, Peleg, Orit, Layer, Ryan
School: University of Colorado at Boulder
Department: Computer Science
School Location: United States -- Colorado
Source: DAI-B 82/2(E), Dissertation Abstracts International
Subjects: Bioinformatics, Toxicology, Genetics
Keywords: Computational toxicology, Mechanistic inference, Nascent transcription, Toxicogenomics, Transcription factor
Publication Number: 27996986
ISBN: 9798664761986
Copyright © 2021 ProQuest LLC. All rights reserved. Terms and Conditions Privacy Policy Cookie Policy