NEW YORK — Polygenic transcriptome risk scores may be better at gauging chronic obstructive pulmonary disease susceptibility across human ancestry groups than polygenic risk scores, a new study has found.
COPD affects about 16 million people in the US and is typically diagnosed through two measures of lung function: forced expiratory volume in one second (FEV1) and the ratio of FEV1 to forced vital capacity. While smoking is a major risk factor for COPD, genetic variation also influences susceptibility.
Previous genome-wide association studies have linked a number of genetic variants to COPD risk, but those studies have drawn largely on populations of European ancestry. Analyses have indicated that polygenic risk scores do not perform as well in target populations that differ from the discovery population.
Researchers from the University of Virginia and elsewhere instead examined whether polygenic transcriptome risk scores (PTRS) could improve cross-ancestry disease risk prediction for COPD over polygenic risk scores (PRS). In previous work, UVa's Ani Manichaikul and colleagues proposed that PTRS developed using multi-SNP predictors of gene expression could improve disease risk predictions across ancestries. Such scores, they said, would focus on the underlying biological mechanisms of a trait or disease, which are shared across ancestries.
"Given that a substantial proportion of GWAS variants demonstrate gene regulation effects, constructing risk scores built on expression quantitative-trait locus (eQTL) variants presents a promising path toward incorporating biological information in genetic prediction," Manichaikul and colleagues wrote in a paper appearing Tuesday in the American Journal of Human Genetics.
For their study, they applied this idea to estimating COPD risk. They generated both PRS and PTRS based on summary statistics from a large GWAS of pulmonary function — both FEV1 and FEV1 to forced vital capacity — and of moderate-to-severe COPD and severe COPD in individuals of European ancestry from the UK Biobank. For the PTRS in particular, they also relied on the PrediXcan approach that combines GWAS data with reference eQTL data.
They then selected the best risk score for each model using the multi-ethnic Trans-Omics for Precision Medicine (TOPMed) population and family-based cohorts and tested them in multi-ethnic TOPMed COPD-enriched cohorts.
Overall, the PTRS had greater portability than the PRS, the researchers found.
The PRS models, they noted, had decreased performance when applied to African Americans, as compared to non-Hispanic whites, dropping from an odds ratio of 1.57 for moderate-to-severe COPD for all participants to 1.24.
The PTRS models, meanwhile, retained their performance and had high portability scores.
Though the PTRS exhibited a weaker association with COPD than the PRS among individuals of European ancestry, it had a stronger association with disease than the PRS among African Americans, especially among heavy smokers for moderate-to-severe COPD. Smoking status and smoking history, the researchers wrote, also has a great effect on COPD risk.
Further, the researchers noted the two score types were not highly correlated, suggesting they could be combined to further improve disease risk predictions.
"While we have demonstrated the value of integrative approaches leveraging eQTLs toward improvement of cross-ancestry portability of risk scores, we further emphasize that constructing more portable risk scores represents just one line of investigation toward achieving equity in risk prediction and personalized medicine," Manichaikul and colleagues wrote. "Ultimately, a crucial step toward improving performance of genomic risk prediction for non-European-ancestry groups will be to increase the diversity of participants included and analyzed in genetic studies."