Stratifying Lung Adenocarcinoma Risk with Multi-ancestry Polygenic Risk Scores in East Asian Never-Smokers

Abstract

Background: Lung adenocarcinoma (LUAD) in never-smokers is a major public health burden, especially among East Asian women. Polygenic risk scores (PRSs) are promising for risk stratification but are primarily developed in European-ancestry populations. We aimed to develop and validate single- and multi-ancestry PRSs for East Asian never-smokers to improve LUAD risk prediction.

Methods: PRSs were developed using genome-wide association study summary statistics from East Asian (8,002 cases; 20,782 controls) and European (2,058 cases; 5,575 controls) populations. Single-ancestry models included PRS-25, PRS-CT, and LDpred2; multi-ancestry models included LDpred2+PRS-EUR128, PRS-CSx, and CT-SLEB. Performance was evaluated in independent East Asian data from the Female Lung Cancer Consortium (FLCCA) and externally validated in the Nanjing Lung Cancer Cohort (NJLCC). We assessed predictive accuracy via AUC, with 10-year and (age 30-80) absolute risks estimates.

Results: The best multi-ancestry PRS, using East Asian and European data via CT-SLEB (clumping and thresholding, super learning, empirical Bayes), outperformed the best East Asian-only PRS (LDpred2; AUC=0.629, 95% CI:0.618,0.641), achieving an AUC of 0.640 (95% CI:0.629,0.653) and odds ratio of 1.71 (95% CI:1.61,1.82) per SD increase. NJLCC Validation confirmed robust performance (AUC =0.649, 95% CI: 0.623, 0.676). The top 20% PRS group had a 3.92-fold higher LUAD risk than the bottom 20%. Further, the top 5% PRS group reached a 6.69% lifetime absolute risk. Notably, this group reached the average population 10-year LUAD risk at age 50 (0.42%) by age 41, nine years earlier.

Conclusions: Multi-ancestry PRS approaches enhance LUAD risk stratification in East Asian never-smokers, with consistent external validation, suggesting future clinical utility.

Keywords: East Asian never smokers; Genome-wide association studies; Lifetime absolute risk; Lung adenocarcinoma; Polygenic risk scores.

Figure 1.. Overview of data structure, polygenic risk score (PRS) development, validation and application.

Summary statistics from East Asian genome-wide association studies (GWAS) were used to develop single-ancestry PRS using methods such as a simple PRS constructed using 25 SNPs that have previously reached genome-wide significance (i.e., P<5×10⁻⁸) (PRS-25), a PRS using the clumping and thresholding (CT) method (PRS-CT) incorporating 8 SNPs, and a PRS using a genome-wide Bayesian-based approach, LDpred2 (LDpred2 PRS) incorporating close to a million SNPs. For the multi-ancestry PRS development, we also used summary statistics from European (EUR) GWAS, applying the PRS-CSx method that leveraged genome-wide association summary statistics for close to a million SNPs with a Bayesian continuous shrinkage prior to model SNP effect sizes across populations, as well as CT-SLEB method, which enhances the standard CT methods with a two-dimensional approach to select SNPS for East Asian PRS construction by incorporating over 2 million SNPs. Tuning and validation of each PRS was conducted in an independent East Asian individual-level data. Relative risk per PRS quantile was calculated as an odds ratio (OR) with the middle quantile (40^th to 60^th percentile) set as the reference, and the area under the receiver operating curve (AUC) was estimated for each PRS. CT-SLEB PRS was used to estimate 10-year and lifetime cumulative absolute risk, and PRS-CT and PRS-LDpred2 were used for sample size projection.

Figure 2.. Relative risk estimated for quantiles of each polygenic risk score (PRS) and lung adenocarcinoma in the validation dataset of women with East Asian ancestry, treating the 40^th to 60^th percentile as the references.

Odds ratios of PRS per standard deviation (SD) and 95% confidence intervals are shown for (A) the single-ancestry 25 SNP polygenic risk score, PRS-25 (A), (B) Clumping and thresholding method, PRS-CT, (C) Bayesian-based genome-wide approach, LDpred2 PRS, and (D) multi-ancestry approach, CT-SLEB.

Figure 3.. Association between polygenic risk score (PRS) and lung adenocarcinoma by age groups.

(A) Odds ratios (ORs) per standard deviation (SD) of the PRS and 95% confidence intervals, and (B) ORs for individuals in the upper 90th percentile of the PRS.

Figure 4.. Lifetime cumulative and 10-year absolute risk of developing lung adenocarcinoma.

(A) Lifetime (age 30–80) cumulative risk and (B) 10-year absolute risk of developing lung adenocarcinoma in never-smoking women in East Asia by percentiles of the CT-SLEB polygenic risk score (PRS). Absolute risks were calculated using the iCARE package, based on Taiwan’s age-specific incidence and mortality data, and the PRS relative risks, as described in the Methods section.

Figure 5.. Projected area under the receiver operating characteristic curve (AUC) of polygenic risk scores (PRS) built using genome-wide association studies (GWAS) with varying sample sizes for lung adenocarcinoma in never-smoking East Asian women.

(A) AUC values for PRS-CT and LDpred2 PRS with case-to-control ratios of 1:1 and 1:10. (B) AUC values for LDpred2 PRS across case-to-control ratios of 1:1, 1:2, 1:3, 1:4, 1:5, and 1:10