The increasing evidence of pleiotropy between complex traits and the increasing functional annotation data call for novel statistical methods to effectively analyze multiple GWAS data sets and functional annotation data simultaneously. Statistical methods to investigate pleiotropy have been actively researched (reviewed in [24] and [25]), for example, using linear mixed models [26], [27] or the conditional False Discovery Rate (FDR) approach [17], [28]. However, these methods do not allow use of functional annotation data for prioritization of GWAS results. On the other hand, various statistical methods have been proposed to make use of functionally annotated SNPs in recent years (reviewed in [29], [30], and [31]). For example, GSEA [32] identifies potentially important pathways in which target genes of risk-associated SNPs are involved while RegulomeDB [33] allows nucleotide-level annotations of risk-associated SNPs, especially for those located in non-coding regions. Stratified FDR methods have been applied to incorporate annotation into GWAS data analysis [20]. However, each of these methods was designed for the analysis of a single phenotype and hence, they do not use functional annotation data fully efficiently for the genetic
