We considered two broad approaches to GSA of SNP data: The commonly used 1-step method in which all SNPs in a gene set are used without consideration of gene-level effects, and a 2-step method where SNPs in each gene are first used to evaluate association with the gene and then gene effects are aggregated to test association with the gene set.
For the 1-step GSA we used Fisher's method to combine SNP p-values, while for the 2-step GSA we ran principal component analysis for genes followed by Fisher's p-value combination.
Because of nonindependence of SNPs, permutations must be used to assess pathway significance, particularly for the 1-step approach. In the 2-step approach, when effects of SNPs in a gene are modeled jointly, the effect of SNP
dependence on the p-value combination step is reduced, allowing for a quick screening of pathway effects without permutations.
Both types of approaches were used to test for association between KEGG pathways and alcohol dependence using data from the Study of Addiction: Genetics and Environment (SAGE).
No significant pathway associations were detected after correction for multiple testing. With the 2-step approach the most significant pathway was "Synthesis and degradation of ketone bodies" (uncorrected p<0.001; p=0.016 with the 1-step approach).
Read Full Abstract (PDF)
