Asessing statistical significance in predictive GWA studies


Testing individual SNPs, as well as arbitrarily large groups of SNPs in GWA studies, using a joint model of all SNPs. The method controls the FWER, and provides an automatic, data-driven refinement of the SNP clusters to smaller groups or single markers.


hierGWAS is a package designed to assess statistical significance in GWA studies, using a hierarchical approach.

There are 4 functions provided: cluster.snp, multisplit, test.hierarchy and compute.r2. cluster.snp performs the hierarchical clustering of the SNPs, while multisplit runs multiple penalized regressions on repeated random subsamples. These 2 functions need to be executed before test.hierarchy, which does the hierarchical testing, though the order in which the 2 functions are executed does not matter. test.hierarchy provides the final output of the method: a list of SNP groups or individual SNPs, along with their corresponding p-values. Finally, compute.r2 computes the explained variance of an arbitrary group of SNPs, of any size. This group can encompass all SNPs, SNPs belonging to a certain chromosome, or an individual SNP.


Laura Buzdugan


Buzdugan, L. et al. (2015), Assessing statistical significance in predictive genome-wide association studies (unpublished)

comments powered by Disqus