Dissertation Defence: Identifying Biological Predictors of Radiation-Induced Lung Injury for Personalized Radiation Therapy

Mitchell Peter Wiebe, supervised by Dr. Christina Haston, will defend their dissertation titled “Identifying Biological Predictors of Radiation-Induced Lung Injury for Personalized Radiation Therapy” in partial fulfillment of the requirements for the degree of Doctor of Philosophy in Medical Physics.

An abstract for Mitchell Peter Wiebe’s dissertation is included below.

Examinations are open to all members of the campus community as well as the general public. This examination will be offered in hybrid format. Registration is not required to attend in person; however, please email christina.haston@ubc.ca to receive the Zoom link for this exam.

Abstract

Radiation therapy (RT) is foundational for lung cancer treatment, but its effectiveness is limited by the risk of radiation-induced lung injury (RILI), including radiation pneumonitis (RP) that may lead to fibrosis. Current clinical estimates of toxicity risk primarily rely on dose-volume histogram (DVH) metrics, such as mean lung dose and volume of the lung receiving ≥ 20 Gy, which describe treatment, but fail to account for inter-patient variability in radiation response. As a result, patients receiving similar RT treatments can experience markedly different toxicity outcomes. The overall goal of this dissertation is to investigate RILI across pre-clinical and clinical settings to identify biological predictors that may complement conventional dose metrics for RILI prediction.

First, Raman spectroscopy was used to investigate metabolic signatures associated with RILI in murine lung tissue. Spectra collected from irradiated and non-irradiated mice were processed and classified with either spectral decomposition techniques coupled with supervised machine learning or through end-to-end classification based on full spectra. Metabolic features were identified that differentiated normal, RP, and fibrotic tissue states, highlighting associations between collagen-related metabolites and fibrosis, as well as inflammatory and immune-related metabolites associated with RP.

Second, pre-treatment circulating blood traits from the Mouse Phenome Database were analyzed to evaluate systemic predictors of radiation response. Several immune and lipid traits were associated with later time to distress and/or extent of fibrosis following irradiation, and supervised learning models demonstrated their potential predictive utility.

Finally, predictive models of RP were developed in a clinical cohort of lung cancer patients receiving RT by integrating autosome-wide genetic variants with clinical, dosimetric, and imaging descriptors. Multi-modal models incorporating genetic variants demonstrated improved predictive performance compared with models based on conventional dose metrics alone, highlighting the contribution of genetic variation to individual RP susceptibility.

Collectively, these findings demonstrate that biological variability contributes meaningfully to RILI susceptibility across multiple biological scales. Incorporating biologically informed predictors may improve toxicity risk stratification and support the development of more personalized and biologically guided RT treatments, ultimately contributing to improved patient outcomes.