NEW YORK — Transcriptome imputation (TI) models derived from individuals of European ancestry can uncover new genetic associations with certain traits in studies of non-European ancestry individuals, a new analysis has found. The researchers added, though, that models derived from more closely matched populations could lead to additional findings.
TI allows researchers to infer tissue-specific gene expression from SNP and associated phenotyping data by using publicly available and well-curated expression quantitative trait locus (eQTL) panels. However, most of these panels and prediction models rely on data from individuals of European ancestry, and it is unclear how well the approach works in other ancestries, according to researchers from the Icahn School of Medicine at Mount Sinai and elsewhere.
In a new study appearing Tuesday in the American Journal of Human Genetics, the researchers applied the PrediXcan approach to impute genetically regulated gene expression in more than 50,000 individuals from the Population Architecture using Genomics and Epidemiology (PAGE) cohort, which includes several different ancestries. Through this, they uncovered more than 100 novel associations with two dozen cardiometabolic traits and uncovered height and body mass-linked associations in the PAGE group that could not be detected in UK Biobank participants of European ancestry.
"Although we and others have demonstrated that TI prediction models have cross-ancestry applicability, it is likely that access to eQTL reference panels derived from matched populations will substantially improve prediction accuracy and that release of non-EA-derived LD matrices will radically improve the applicability of existing TI prediction models," senior author Laura Huckins from Mount Sinai and her colleagues wrote in their paper.
The researchers applied the PrediXcan transcriptomic imputation approach to the PAGE cohort, which includes individuals of African American, Hispanic or Latino, Asian, and Hawaiian ancestries, to uncover 102 novel associations with 25 cardiometabolic traits in relevant tissues. These traits included lipid traits like cholesterol level, lifestyles traits like smoking, inflammatory traits like C-reactive protein level, and more.
In all, they uncovered 1,113 tissue-specific genetically regulated gene expression events associated with one of those traits, including 102 novel gene-trait associations.
The researchers then compared transcriptomic imputation findings from the PAGE cohort to a similar-sized cohort from the UK Biobank that was of white European ancestry. For this, they focused on two anthropometric traits — height and BMI — that were well characterized in both cohorts.
They uncovered substantial overlap between associated genes uncovered in the PAGE and UK Biobank. For instance, in whole blood, 93 genes were associated with height in the UK Biobank and 10 of these were also significant in the PAGE cohort, more than would be expected by chance. At the same time, 27 genes were significantly associated with height in PAGE, 10 of which were also significant in the UK Biobank. Similar patterns were also found for BMI.
Still, the researchers estimated that about 65 percent of the signals found in PAGE were not significantly associated in the UK Biobank, which they said underscores the need for diverse samples.
These and other findings suggested that even though transcriptomic imputation prediction models can work across ancestries, having eQTL reference panels derived from matched populations would boost their accuracy.
"Moreover, there is an ethical imperative to improve diversity in functional genomics and transcriptomics studies; opportunities to participate in groundbreaking research approaches should be equally available regardless of race or ethnicity, and the scientific insights obtained should be applicable and accessible to all," Huckins and colleagues added.