SFDR (Stratified False Discovery Rate) Software Documentation

False Discover Rate (FDR) method provides more powerful multiple hypothesis testing criteria than the conventional Family-Wise Error Rate (FWER) control method (Benjamini & Hochberg, 1995). FDR method, therefore, is well-suited for Genome-Wide Association (GWA) study of several hundred thousands of SNPs. FDR control is performed collectively to the p-values from all hypotheses, and the critical point for p-values varies for a fixed FDR control level depending on the underlying distribution of true alternative hypothesis. The FDR method can be also performed by obtaining FDR adjusted p-value, i.e. q-value for each hypothesis testing and comparing q-values with FDR control level directly (Storey,2002). Stratified False Discovery Rate (SFDR) method performs FDR control separately for each stratum which has been decided by prior information or some characteristic of data (Sun et al., 2006). Weighted False Discovery Rate (WFDR) method assigns specific weight for each hypothesis and performs FDR control on weighted p-values (Roeder et al., 2006). These two FDR methods that incorporate prior information are especially suitable for the prior information of Genome-Wide Linkage study results used for the analysis of GWA study. For SFDR method, the SNPs are divided into two strata using a threshold value for linkage score. If the corresponding linkage score at the location of a SNP is above the threshold, the SNP is assigned to the stratum 1 (high linkage stratum). Otherwise, the SNP is assigned to the stratum 2 (low linkage stratum). For WFDR method, weight of SNP i is computed as the wi=exp(B∙Zi)/v where N i