Most GWAS loci fall in regions that control gene expression rather than in the genes themselves, and linking those variants to eQTL and sQTL signals has typically hit a discovery limit after a few thousand samples. A new study published in Science suggests that limit was really a limit of dataset size.
The team analyzed 14,324 RNA sequencing samples, including 6,454 whole blood and 1,291 lung samples, from the NHLBI-sponsored Trans-Omics for Precision Medicine (TOPMed) program. After e/sQTL analysis, they mapped tens of thousands of signals at high resolution. Because TOPMed cohorts are more genetically diverse than most prior studies, roughly one in six signals found in whole blood came from variants that are rare or absent in people of European ancestry, and just 1.3% of those appear in the four large eQTL resources built mostly from European samples.
“The more samples you add, the more e/sQTL signals that you detect,” said co-study leader Steve Parker from the University of Michigan. “I think NHLBI has invested very smartly in TOPMed. Although you saturate discovery of main effects after a few thousand samples… the number of independent e/sQTL signals just keeps on growing and growing, which is really cool to see.”
Fine-mapping GWAS signals from the UK Biobank showed that more eQTL signals also meant more overlap with GWAS findings, and the extra signals weren’t redundant: most genes have several independent regulatory switches, and the second, third, and fourth are harder for smaller studies to detect. Of the 10,611 GWAS signals the team connected to a specific gene, roughly two-thirds involved one of these secondary switches.
“Genetic regulation of the expression levels of one part of a biological unit’s machinery, such as a regulator of a cellular stress response, can affect the expression levels of other parts of the same biological machinery,” explained co-author Laura Scott. Parker added that “genetic evidence of drug targets is a huge predictor of clinical success for therapeutics,” more than doubling the rate of success, making the link between genetic variation and target gene expression especially valuable.
Co-author Francois Aguet noted that “another powerful aspect of larger-scale studies is the ability to identify QTL for rarer variants, which tend to have larger effects. As further such studies are conducted across more tissues and cell types, we will likely find a lot more of these effects.”
The findings also point to the value of expanding beyond whole blood and improving ancestral diversity. “In TOPMed, we still have many more distinct signals to discover at most genes, even in whole blood, which will almost certainly provide still more links between GWAS loci and their likely mechanisms,” noted co-author Laura Raffield. “Also, many non-European ancestry populations are still underrepresented in our current dataset… important ancestry-differentiated signals may be missed, even in blood, and certainly in other tissues. These issues will continue to be addressed in future work from TOPMed.”
“Science moves forward with bigger and better data, always,” Parker said. “It’s the equivalent of getting a fancy new microscope except we’re not seeing things in higher resolution in a microscopy image, we’re seeing things in much higher resolution from a genetic association with gene expression perspective.”
The team analyzed 14,324 RNA sequencing samples, including 6,454 whole blood and 1,291 lung samples, from the NHLBI-sponsored Trans-Omics for Precision Medicine (TOPMed) program. After e/sQTL analysis, they mapped tens of thousands of signals at high resolution. Because TOPMed cohorts are more genetically diverse than most prior studies, roughly one in six signals found in whole blood came from variants that are rare or absent in people of European ancestry, and just 1.3% of those appear in the four large eQTL resources built mostly from European samples.
“The more samples you add, the more e/sQTL signals that you detect,” said co-study leader Steve Parker from the University of Michigan. “I think NHLBI has invested very smartly in TOPMed. Although you saturate discovery of main effects after a few thousand samples… the number of independent e/sQTL signals just keeps on growing and growing, which is really cool to see.”
Fine-mapping GWAS signals from the UK Biobank showed that more eQTL signals also meant more overlap with GWAS findings, and the extra signals weren’t redundant: most genes have several independent regulatory switches, and the second, third, and fourth are harder for smaller studies to detect. Of the 10,611 GWAS signals the team connected to a specific gene, roughly two-thirds involved one of these secondary switches.
“Genetic regulation of the expression levels of one part of a biological unit’s machinery, such as a regulator of a cellular stress response, can affect the expression levels of other parts of the same biological machinery,” explained co-author Laura Scott. Parker added that “genetic evidence of drug targets is a huge predictor of clinical success for therapeutics,” more than doubling the rate of success, making the link between genetic variation and target gene expression especially valuable.
Co-author Francois Aguet noted that “another powerful aspect of larger-scale studies is the ability to identify QTL for rarer variants, which tend to have larger effects. As further such studies are conducted across more tissues and cell types, we will likely find a lot more of these effects.”
The findings also point to the value of expanding beyond whole blood and improving ancestral diversity. “In TOPMed, we still have many more distinct signals to discover at most genes, even in whole blood, which will almost certainly provide still more links between GWAS loci and their likely mechanisms,” noted co-author Laura Raffield. “Also, many non-European ancestry populations are still underrepresented in our current dataset… important ancestry-differentiated signals may be missed, even in blood, and certainly in other tissues. These issues will continue to be addressed in future work from TOPMed.”
“Science moves forward with bigger and better data, always,” Parker said. “It’s the equivalent of getting a fancy new microscope except we’re not seeing things in higher resolution in a microscopy image, we’re seeing things in much higher resolution from a genetic association with gene expression perspective.”