Back to search

University of Alabama Birmingham

Application of Machine Learning and Network Approaches to Prioritize Safe and Efficacious Drug Repurposing Candidates

Abstract

dc:description.abstract

Only ten percent of drugs in clinical trials are ultimately approved by the Food and Drug Administration (FDA). With many diseases and disorders needing therapeutic options, computational drug repurposing methodologies are great alternatives to some drug discovery methods due to quicker FDA approval and lower cost. We sought to gain insights to improve the identification of drug repurposing candidates that will be safe and effective to increase the success of computational drug repurposing approaches. In regards to identifying safe drug candidates, pharmacovigilance studies of adverse event case reports identified that women are more likely to experience an adverse event, and men are more likely to experience severe adverse events. We investigated drugs associated with sex-biased adverse events (SBAEs) and identified if there are commonly utilized drug targets and metabolism enzymes. Then, we learned that these common drug targets and metabolism enzymes were more likely to have sex-biased genomic features (i.e., gene expression and gene-regulatory network properties) that could be used in the future to select drug repurposing candidates less likely to cause SBAEs. Prioritizing effective drug candidates is especially critical for cancers, as it is estimated that only five percent of oncology drugs in clinical trials are FDA approved. To address this challenge, previous cancer studies applied and validated transcriptomic signature reversion, a computational drug repurposing approach. However, it was unclear if different disease-associated gene signature methods for signature reversion resulted in different drug candidates. Therefore, for four low-survival cancers (i.e., glioblastoma [GBM], liver hepatocellular carcinoma [LIHC], lung adenocarcinoma [LUAD], and pancreatic adenocarcinoma [PAAD]), we applied three disease-associated gene signature methods for signature reversion: differential gene expression analysis (DESeq2 & limma) and transfer learning. Across the cancers, we identified several drug candidates from all three approaches that decreased cell growth in the PRISM cancer drug screen and were previously in cancer clinical trials. Overall, we provide evidence of sex-biased genomic features of SBAE-associated drug targets and metabolism enzymes and the benefit of multiple disease-associated gene signatures for transcriptomic signature reversion drug repurposing. These findings will be important for future sex-aware and cancer transcriptomic signature reversion drug repurposing applications.

Degree

thesis:*
Level thesis:degree_level
Dissertation

Author and committee

dc:creator, dc:contributor.*
Author dc:creator
  • Fisher, Jennifer L
Contributors dc:contributor
  • Brittany N Lasseinge
  • Anita Hjelmeland
  • Lara Ianov
  • Marek Napierala
  • Akinyemi I Ojesina
  • Elizabeth A Worthey

Subjects

dc:subject × 2

Identifiers

dc:identifier.*

Chain of custody

source
Harvested from
University of Alabama Birmingham
Base URL
digitalcommons.library.uab.edu/do/oai/
Last updated
2026-11-19
Source record
OAI-PMH GetRecord
citation

Fisher, Jennifer L. Application of Machine Learning and Network Approaches to Prioritize Safe and Efficacious Drug Repurposing Candidates. Dissertation thesis, https://digitalcommons.library.uab.edu/etd-collection/430