Back to search

University of British Columbia

Pharmacogenomic and machine learning insights for the prediction of l-asparaginase-induced hypersensitivity in pediatric cancer

Abstract

dc:description

Hypersensitivity reactions to l-asparaginase pose a major limitation in pediatric acute lymphoblastic leukemia (ALL) therapy, often necessitating treatment modifications that jeopardize remission and survival. This dissertation investigates the genetic underpinnings of these reactions and explores strategies for identifying high-risk patients. A systematic review of studies examining the genetic associations of l-asparaginase-induced hypersensitivity in human case-control studies confirmed a consistent, albeit modest, effect of HLA-DRB1*07:01 (p = 3.51 x 10⁻¹⁸, OR = 1.96 [1.68–2.28], I² = 18.42%) on hypersensitivity risk. By contrast, genome-wide association studies produced heterogeneous findings. These observations emphasize the importance of robust phenotyping and ancestry-aware study designs. In a genome-wide associations study of l-asparaginase-induced hypersensitivity in a large discovery cohort (n = 926) of pediatric cancer patients, large effect variants (OR = 3.9-8.5, p < 10⁻⁵) were identified within regulatory regions of genes (CYP1B1, OPLAH, SORCS2, SEC16B) involved in amino acid stress response, endoplasmic reticulum stress, and l-asparaginase clearance. These findings were then validated in an independent replication cohort (n = 180). When top-risk variants were incorporated into a genetic risk score, carriers of two or more risk alleles displayed a 25-fold increased risk of experiencing hypersensitivity (OR = 25.2 [7.4-86.2], p = 1.0 x 10⁻¹⁰), underscoring the power of polygenic models to enhance individual risk stratification. Predictive modeling revealed that incorporating multiple polygenic risk scores from diverse studies, along with clinical variables and drug usage data, improved risk prediction by 36% over single-variant models. The most comprehensive machine learning frameworks (neural networks, support vector machines, logistic regression) achieved area under the curve (AUC) scores > 0.70 in an independent replication cohort, with significant differences for the prediction of cases of hypersensitivity compared to controls (OR = 3.4 - 4.5, p < 3.86 x 10⁻⁴), demonstrating discrimination between patients who developed hypersensitivity and those who did not. These findings highlight the complex genetic architecture of l-asparaginase hypersensitivity and the potential of comprehensive risk scores in refining patient stratification. By integrating genetic, clinical, and pharmacological data, this work provides a framework for pharmacogenetic testing to preserve treatment efficacy while minimizing adverse events, advancing precision medicine in pediatric ALL care.

Degree

thesis:*
Name thesis:degree_name
Doctor of Philosophy - PhD
Level thesis:degree_level
doctoral
Discipline thesis:degree_discipline
Medical Genetics
Grantor dc:publisher
University of British Columbia
Year dc:date
2025

Author and committee

dc:creator, dc:contributor.*
Author dc:creator
  • Anderson, Spencer
Contributors dc:contributor
  • Carleton, Bruce C.
  • Ross, Colin J.

Rights

dc:rights
Statement dc:rights
  • Attribution-NonCommercial-NoDerivatives 4.0 International
Language dc:language
eng

Identifiers

dc:identifier.*
Handle dc:identifier
http://hdl.handle.net/2429/90642

Chain of custody

source
Harvested from
University of British Columbia
Base URL
circle.library.ubc.ca/oai/request
Last updated
2026-11-19
Source record
OAI-PMH GetRecord
related terms
citation

Anderson, Spencer. Pharmacogenomic and machine learning insights for the prediction of l-asparaginase-induced hypersensitivity in pediatric cancer. doctoral thesis, University of British Columbia, 2025. http://hdl.handle.net/2429/90642