11 / 2026-01-29 09:02:29

Development and validation of a multi-ancestry and multi-trait polygenic risk score for lung cancer

polygenic risk score,multi-ancestry,multi-trait,lung cancer,genome-wide association study

前沿论坛 > 基因组学与医药健康

Polygenic risk score (PRS) can enhance risk stratification and help identify high-risk populations. However, the ethnic imbalance in genome-wide association studies (GWAS) for lung cancer leads to the poor accuracy of existing monoracial PRSs in non-European populations. Here, we performed a multi-ancestry GWAS meta-analysis using GWAS summary statistics of lung cancer (76,953 cases and 1,886,372 controls) across diverse ancestries. We then developed a multi-ancestry PRS of lung cancer (PRSMA ) using PRS-CSx. Moreover, to enhance the performance of PRSMA , we constructed a multi-trait PRS of lung cancer (PRSMT ), using categorical boosting (CatBoost) based on 32 cross-trait PRSs for lung cancer across three ancestries. Finally, a multi-ancestry and multi-trait PRS, termed PRSMAMT , was generated by integrating PRSMA  and PRSMT . We validated PRSMAMT  in three independent multi-ancestry cohorts [International Lung Cancer OncoArray Consortium (OncoArray), Transdisciplinary Research Into Cancer of the Lung (TRICL) and All of Us (AoU)]. Compared to previously reported loci from earlier studies, 69 new independent genome-wide significant loci were identified based on the multi-ancestry GWAS meta-analysis, including 2 located in new cytoband regions. Compared to 32 published lung cancer PRSs, PRSMA  ranked first on average in the overall population.  Superior discrimination of PRSMAMT  was observed in European, Asian, and African populations, with average area under the receiver operating characteristic curves (AUCs) of 0.600, 0.590, and 0.609, respectively. Furthermore, PRSMAMT  enhanced risk stratification capability and identified approximately 10% additional lung cancer cases beyond the Prostate, Lung, Colorectal and Ovarian (PLCO_m2012) model in UK Biobank. Individuals with both high PLCO_m2012 scores and high genetic risk exhibited a 12.64-fold cumulative risk of lung cancer (95% confidence interval: 11.60-13.78) compared with the reference population with both low scores. Our results demonstrate that PRSMAMT  can effectively improve lung cancer risk prediction across ancestries and may facilitate precision prevention and targeted interventions.