Grant: Master's Award: Canada Graduate Scholarships

Lung function, measured by spirometry, is a key clinical indicator of respiratory diseases like asthma and chronic obstructive pulmonary disease, which collectively cost the Canadian healthcare system $3.6 billion annually. Previous work in our lab has established a polygenic score (PGS) for lung function—a score estimating the cumulative effect of many genetic variants on an individual's phenotype—but this explains only 5% of the variance in lung function, as is the case for many published PGSs. This highlights the critical role of non-genetic factors, such as environmental exposures, in shaping respiratory health. Our study aims to predict individuals’ lung function by integrating genetic and environmental data, revealing the complex, cumulative interactions between these factors that mediate lung function. Using data from the CHILD Cohort Study, which includes prenatal and postnatal variables for 1,510 mother-infant pairs with infant lung function measurements, we will develop a machine learning (ML) model capable of capturing intricate, non-linear relationships. However, while complex ML models can achieve superior predictive accuracy, their “black box” nature limits interpretability, making it challenging to translate insights into actionable recommendations. To overcome this, we will apply explainable artificial intelligence (AI) techniques inspired by psychological heuristics, such as counterfactual reasoning, to elucidate how variables interact and influence lung function outcomes. This study will provide a more wholistic understanding of how genetics and the environment impact lung function, while introducing strategies of how we can modify environmental variables to make lifestyle changes for health.