Introduction to GWAS2Genes
About
GWAS2Genes is a database for linking common variants of GWAS loci to which genes are affected and in which direction by using eQTLs and SMR. Below is the abstract for our paper:
Genome-wide association studies (GWASs) have identified a multitude of genetic loci involved with traits and diseases. However, it is often unclear which genes are affected in such loci and whether the associated genetic variants lead to increased or decreased gene function. To mitigate this, we integrated associations of common genetic variants in 57 GWASs with 24 studies of expression quantitative trait loci (eQTLs) from a broad range of tissues by using a Mendelian randomization approach. We discovered a total of 3,484 instances of gene-trait-associated changes in expression at a false-discovery rate < 0.05. These genes were often not closest to the genetic variant and were primarily identified in eQTLs derived from pathophysiologically relevant tissues. For instance, genes with expression changes associated with lipid traits were mostly identified in the liver, and those associated with cardiovascular disease were identified in arterial tissue. The affected genes additionally point to biological processes implicated in the interrogated traits, such as the interleukin-27 pathway in rheumatoid arthritis. Further, comparing trait-associated gene expression changes across traits suggests that pleiotropy is a widespread phenomenon and points to specific instances of both agonistic and antagonistic pleiotropy. For instance, expression of SNX19 and ABCB9 is positively correlated with both the risk of schizophrenia and educational attainment. To facilitate interpretation, we provide this lexicon of how common trait-associated genetic variants alter gene expression in various tissues as the online database GWAS2Genes.
We also provide a brief overview of these eQTL and GWAS studies
How to use
The GWAS2Gene database have three different components: Firstly, the trait associated gene heatmaps show each gene found to be associated with a trait of interest at a study wide FDR < 5%. Secondly, Manhattan plots that for a given trait shows the strength of association for with genes in either all or a selected tissue. If all tissues are selected here, then the strongest association is shown. Thirdly, detailed results for the individual associations between eQTLs and GWAS are also provided. These can be filtered using the "p_SMR" and "p_HET" cut-offs as detailed below.
To understand the results, two parameters from SMR are particularly imporant: SMR tests for a joint association between the eQTL and GWAS data resulting in the "p_SMR" p-value. SMR further analyzes the profile of association for nearby co-inherited variants in the GWAS and eQTL study to test if the signals are dissimilar in a so-called heterogeneity in dependent instruments (HEIDI) giving the "p_HET" p-value. If that is the case (i.e. the p_HET test is significant), the identified GWAS and eQTL signals are less likely to be driven by the same genetic variant, and the overlap can be thus be incidental and driven by linkage. In our paper focussed on the associations that had a study-wide significant p_SMR and didn't show significant heterogeneity (p_HET>0.05). We indicate which associations failed the heterogeneity test in the heatmaps, for the detailed results view, the p_SMR and p_HET cutoffs are user settable.
If you have further qustions about using the page please contact Mads Hauberg: hauberg@biomed.au.dk; Wen Zhang: zhang.wen81@gmail.com (website), or Panos Roussos: panagiotis.roussos@mssm.edu for more information.
How to cite
If you use the GWAS2Genes database in an article, then please cite the original GWAS(s) that your article involves as well as well as our paper:
Large-Scale Identification of Common Trait and Disease Variants Affecting Gene Expression. Mads Engel Hauberg, Wen Zhang, Claudia Giambartolomei, Oscar Franzén, David L. Morris, Timothy J. Vyse, Arno Ruusalepp, CommonMind Consortium, Pamela Sklar, Eric E. Schadt, Johan L.M. Björkegren, Panos Roussos. DOI: http://dx.doi.org/10.1016/j.ajhg.2017.04.016
Download
You can download the complete database as flat files here. (515MB). The main results file is too large to be opened in Excel or similar programs, so this only relevant if you can used more specialized tools like for instance "R".
