to detect small effects is linked primarily to sample size (given a fixed alpha level). However, power may also depend on the nature of the genotype-phenotype relation 14, 15. Most consortia focus initially on individual SNPs showing the strongest evidence for association. The TAG consortium with the largest sample yet of 74,053 individuals identified 130 SNPs (tagging the 15q25 locus) that passed the genome-wide threshold of 5×10-8 in the quantity smoked analysis. Subsequently, 1025 SNPs that passed the significance threshold of 10-4 were included in the follow-up SNP-based analyses in the combined TAG, ENGAGE12 and the Oxford-GlaxoSmithKline11 sample. We note that the remaining data (summary statistics for up to ∼2.5 million SNPs in each analysis) have remained largely unexploited. Set-based tests (with a gene or a biological pathway as the unit of analysis) are more powerful than SNP based tests 14, 16 as they consider jointly the weak effects of multiple SNPs within a gene or biological pathway. Furthermore, by targeting genomic regions rather than individual SNPs, the number of tests drops from millions to thousands, thus alleviating the multiple testing burden.
