Abstract
INTRODUCTION: Lung cancer is a leading cause of cancer-related mortality globally, with distinct epidemiological and genetic patterns in East Asian populations. However, most polygenic risk score (PRS) models have been developed using European-ancestry cohorts, raising concerns about their applicability in non-European populations. MATERIALS AND METHODS: In this study, we systematically evaluated the predictive performance of three PRS approaches in a Chinese lung cancer cohort consisting of 97 cases and 667 controls. We assessed (i) a previously reported 19-SNP PRS developed in Chinese individuals, (ii) genome-wide PRS derived using PRS-CS with East Asian and European GWAS summary statistics, and (iii) PRS-CSx, a cross-population Bayesian framework that integrates summary statistics across ancestries. RESULTS: The 19-SNP PRS demonstrated limited discriminative power in our cohort. In contrast, PRS-CS using East Asian summary statistics showed significant associations with overall lung cancer and specific histological subtypes, particularly NSCLC and LUAD. PRS-CS based on European data yielded weaker performance, underscoring the importance of ancestry matching. Notably, PRS-CSx outperformed single-ancestry models, achieving improved risk stratification for NSCLC and LUAD. However, its predictive performance for LUSC and SCLC remained limited, likely due to sample size constraints and subtype heterogeneity. CONCLUSION: Our findings emphasize the critical role of ancestry-matched data and integrative PRS approaches in enhancing risk prediction in underrepresented populations. PRS-CSx represents a promising tool for lung cancer risk assessment in East Asians, though further validation in larger cohorts are needed to improve generalizability and clinical utility.