LD clumping is a commonly used technique. It provides a reasonable solution to the problem that often you only want the best SNP from a GWAS signal, not all the SNPs in the region.
LD clumping is performed as follows.
--clump-p1)--clump-kb) of the best SNP that exceed another p-value threshold (--clump-p2)--clump-r2) form a clump for the best SNP, and are discarded from being potential future best SNP candidatesThe basic idea is that if a SNP is in LD with the top SNP in the region then we assume that it only appears to have an effect because it is associated with the top SNP.
We can use plink to perform LD clumping. Let’s navigate to the folder that contains our GWAS results
cd results\gwasPerforming clumping on the BMI GWAS
..\..\software\plink_win.exe^
--bfile ..\..\data\genetic\geno_qc^
--clump precomputed\bmi_adjusted.assoc.linear.add.gz^
--clump-p1 5e-8^
--clump-p2 5e-8^
--clump-kb 1000^
--clump-r2 0.01^
--out bmiThe flags being used here are
--clump precomputed/bmi_adjusted.assoc.linear.add.gz specifies the file with the results. All that is required is a column with SNPs (labelled SNP in the header) and a column with P values (labelled P in the header)--clump-p1 5e-8 specifies to only allow best SNPs with this minimum p-value--clump-p2 5e-8 specifies that SNPs to include in the clump list are only retained if they surpass this p-value--clump-kb 1000 specifies the radius around which to look for clumps (1Mb in this case)--clump-r2 0.01 specifies to include SNPs with r-squared greater than 0.01 in the clump for the best SNP.This creates a file called bmi.clumped. It has one line for every clump, with the best SNP specified in the SNP column. The SNPs that were within the radius, had the p2 p-value, and were in LD above the threshold, are listed in the SP2 column.
Now do the same thing with the other GWAS results too.
Let’s first look at the UCSC genome browser. Choose one of the retained SNPs and use it as a search term here
http://genome.ucsc.edu/cgi-bin/hgGateway
Zooming out may help to see more information. The UCSC genome browser has a huge amount of information that is organised in ‘tracks’. A track is a layer of information that is mapped to a genome build. For example, the common SNPs track shows SNP positions within the window that is visible. There is a ‘RefSeq Genes’ track which shows the orientation, structure and position of genes.
Note that the SNP you chose likely is listed in the ‘NHGRI-EBI Catalog of Published Genome-Wide Association Studies’ track. If you click on a SNP there it will tell you information about the publication.
A useful resource for finding out more information about a particular SNP is the Haploreg website.
http://www.broadinstitute.org/mammals/haploreg/haploreg.php
Enter your SNP into this search box. What this database does is it provides information about the SNP that you searched for, for example whether it lies in DNA motifs, which tissues it might be relevant to, etc; but it also shows the same information for SNPs that are in LD with the SNP you specified. This is valuable because it could be that your SNP is not the causal variant, and it is simply in LD with the true causal variant which will have some relevant annotations.