Evans, David M2; Brion, Marie Jo A2; Paternoster, Lavinia2; Kemp, John P2; McMahon, George2; Munafò, Marcus2; Whitfield, John B2; Medland, Sarah E2; Montgomery, Grant W2; Timpson, Nicholas J2; St Pourcain, Beate2; Lawlor, Debbie A2; Martin, Nicholas G2; Dehghan, Abbas2; Hirschhorn, Joel2; Davey Smith, George2; Consortium, GIANT2; Jørgensen, Torben3
1 Department of Public Health, Department of Public Health, Faculty of Health and Medical Sciences, Københavns Universitet2 unknown3 Department of Public Health, Department of Public Health, Faculty of Health and Medical Sciences, Københavns Universitet
It is common practice in genome-wide association studies (GWAS) to focus on the relationship between disease risk and genetic variants one marker at a time. When relevant genes are identified it is often possible to implicate biological intermediates and pathways likely to be involved in disease aetiology. However, single genetic variants typically explain small amounts of disease risk. Our idea is to construct allelic scores that explain greater proportions of the variance in biological intermediates, and subsequently use these scores to data mine GWAS. To investigate the approach's properties, we indexed three biological intermediates where the results of large GWAS meta-analyses were available: body mass index, C-reactive protein and low density lipoprotein levels. We generated allelic scores in the Avon Longitudinal Study of Parents and Children, and in publicly available data from the first Wellcome Trust Case Control Consortium. We compared the explanatory ability of allelic scores in terms of their capacity to proxy for the intermediate of interest, and the extent to which they associated with disease. We found that allelic scores derived from known variants and allelic scores derived from hundreds of thousands of genetic markers explained significant portions of the variance in biological intermediates of interest, and many of these scores showed expected correlations with disease. Genome-wide allelic scores however tended to lack specificity suggesting that they should be used with caution and perhaps only to proxy biological intermediates for which there are no known individual variants. Power calculations confirm the feasibility of extending our strategy to the analysis of tens of thousands of molecular phenotypes in large genome-wide meta-analyses. We conclude that our method represents a simple way in which potentially tens of thousands of molecular phenotypes could be screened for causal relationships with disease without having to expensively measure these variables in individual disease collections.
P L O S Genetics (online), 2013, Vol 9, Issue 10, p. 1-15