ORCID

http://orcid.org/0000-0001-6221-9014

Date of Award

Winter 12-15-2018

Author's School

Graduate School of Arts and Sciences

Author's Department

Biology & Biomedical Sciences (Computational & Systems Biology)

Degree Name

Doctor of Philosophy (PhD)

Degree Type

Dissertation

Abstract

Metabolism can convert drugs to harmful reactive metabolites that conjugate to DNA and off-target proteins. Reactive metabolites are a significant driver of both drug candidate attrition and withdrawal from the market of already approved drugs. Unfortunately, reactive metabolites are difficult to study in vivo, because they are transitory and generally do not circulate. Instead, this work computationally models both metabolism and reactivity. Using deep learning, predictive models were developed for the metabolic formation of quinones and epoxides, which together account for about half of known reactive metabolites. Additionally, an accurate model of DNA and protein reactivity was constructed, which predicts how likely a molecule is to be reactive, and therefore potentially toxic. To connect the metabolism and reactivity models, a system was developed for predicting the exact structures of quinones and epoxides. Finally, using the metabolite structure predictor as a stepping-stone, the quinone formation and epoxidation models were connected to the reactivity model to build an integrated bioactivation model of metabolism and reactivity.

Language

English (en)

Chair and Committee

Sanjay J. Swamidass

Committee Members

Jeremy Buhler, Mark Anastasio, Gregory Bowman, Gary Stormo,

Comments

Permanent URL: https://doi.org/10.7936/2x7h-bq64

Download

Share

COinS