Knowledge-Guided Deep Learning Models of Drug Toxicity Improve Interpretation / Yun Hao.

Format:

Book

Thesis/Dissertation

Author/Creator:

Hao, Yun, author.

Contributor:

University of Pennsylvania. Genomics and Computational Biology, degree granting institution.

Language:

English

Subjects (All):

Bioinformatics.

Genomics and Computational Biology--Penn dissertations.

Penn dissertations--Genomics and Computational Biology.

Local Subjects:

Bioinformatics.

Genomics and Computational Biology--Penn dissertations.

Penn dissertations--Genomics and Computational Biology.

Physical Description:

1 online resource (116 pages)

Contained In:

Dissertations Abstracts International 84-08B.

Place of Publication:

[Philadelphia, Pennsylvania] : University of Pennsylvania, 2022.

Ann Arbor : ProQuest Dissertations & Theses, 2022

Language Note:

English

Summary:

In drug development, a major reason for attrition is the lack of understanding of cellular mechanisms governing drug toxicity. Conventional models are limited by low accuracy and lack of interpretability. Further, they often fail to explain cellular mechanisms underlying structure-toxicity associations. To address these limitations, we developed a series of interpretable in silico models that connect drugs to their toxicity targets and pathways. In Chapter 2, we incorporated target profile as an intermediate connecting structure to toxicity. To accommodate for high-dimensional feature space, we developed a pipeline named TargetTox that can identity subset of predictive features. The features identified by TargetTox accurately predicted binding outcomes for 377 targets and toxicity outcomes for 36 adverse events. We demonstrated that predictive targets tend to be differentially expressed in the tissue of toxicity. We also discovered that predictive targets are enriched for key cellular functions associated with toxicity. Furthermore, we found evidence supporting diagnostic/therapeutic applications of some predictive targets. Our findings highlighted the critical role of predictive targets in cellular mechanisms leading to toxicity. In Chapter 3, we developed DTox, an interpretation framework for knowledge-guided neural networks, which can predict compound response to toxicity assays and infer toxicity pathways of individual compounds. We demonstrate that DTox can achieve the same level of predictive performance as conventional models with a significant improvement in interpretability. Using DTox, we were able to rediscover mechanisms of transcription activation by nuclear receptors, recapitulate cellular activities induced by aromatase inhibitors and PXR agonists, and differentiate distinctive mechanisms leading to HepG2 cytotoxicity. Virtual screening by DTox revealed that compounds with predicted cytotoxicity are at higher risk for clinical hepatic phenotypes. In Chapter 4, we introduce AIDTox, an interpretable deep learning model which incorporates curated knowledge of chemical-gene connections, gene-pathway annotations, and pathway hierarchy. AIDTox can accurately predict cytotoxicity outcomes. It also provides comprehensive explanations of cytotoxicity covering multiple aspects of drug activity including target interaction, metabolism, and elimination. In summary, our work provides a framework for deciphering cellular mechanisms of toxicity in silico.

Notes:

Source: Dissertations Abstracts International, Volume: 84-08, Section: B.

Advisors: Shen, Li; Moore, Jason H.; Committee members: Holmes, John H.; Penning, Trevor M.; Tatonetti, Nicholas P.; Urbanowicz, Ryan J.

Department: Genomics and Computational Biology.

Ph.D. University of Pennsylvania 2022.

Local Notes:

School code: 0175

ISBN:

9798374414363

Access Restriction:

Restricted for use by site license.

This item is not available from ProQuest Dissertations & Theses.