Open Collections

Trinh, Huong Ly

University of British Columbia

Background: Metabolomics offers a promising approach to identifying biomarkers for timely intervention and enhanced screening for individuals at risk of breast cancer. In this nested case-control study, we used self-reported risk factors and prospectively collected blood samples to identify metabolites associated with breast cancer and develop predictive models to assess population risk profiles. Methods: Participants from two Canadian prospective cohorts, BC Generations Project and Alberta’s Tomorrow Project, completed baseline health and lifestyle questionnaires and donated blood samples at enrollment. 593 women who subsequently developed breast cancer cases were matched one-to-one with healthy controls based on cohort, age at blood collection, time of blood collection, and menopause status. Quadrupole time-of-flight mass spectrometry was used for untargeted metabolomics profiling of baseline serum samples. Logistic regression was used to estimate odds ratios corresponding to a one standard deviation increase of metabolite levels after adjusting for matching variables and selected health and lifestyle risk factors. Associations were also assessed for matched pairs with postmenopausal cancer, invasive ductal carcinoma, and hormone receptor-positive cancer. Five classifiers – lasso, random forest (RF), and partial least squares discriminant analysis (PLS-DA), and support vector machines (SVMs) with linear and radial basis functions – were used to develop risk prediction models. Results: 24 metabolites were significantly associated with risk of developing breast cancer, including 13 associated with decreased risk and 11 associated with increased risk. No associations were attenuated by more than 10% when adjusted for risk factors. Among the 24 metabolites, associations with postmenopausal breast cancer, invasive ductal carcinoma, and ER+/PR+ breast cancer were found for 10, 10, and 4 metabolites, respectively. Prediction models using these metabolites had moderate discrimination, with highest AUC of 0.63. Lasso and SVM with linear basis function outperformed the other three algorithms. Conclusion: Overall, our results suggest metabolomics data shows potential as breast cancer biomarkers. More studies are needed to confirm the identities of detected metabolites and validate their associations with breast cancer risk. Future analyses that incorporate larger sample sizes and longitudinal samples will further provide more insights into the metabolic mechanisms surrounding cancer development.

Text

2024-05-03

Attribution-NonCommercial-NoDerivatives 4.0 International

http://hdl.handle.net/2429/88168

Population and Public Health

University of British Columbia

UBCV

http://creativecommons.org/licenses/by-nc-nd/4.0/