Comparing multiple testing correction methods between two softwares for single nucleotide polymorphisms association analyses: Using OPRD1 and diastolic blood pressure in methadone maintenance patients as an example

Multi-testing correction methods are usually applied in more than one single nucleotide polymorphisms (SNPs) of association analyses to determine the significance of results in the face of testing multiple hypotheses. One of the statistical methods applied in these analyses is called false discovery rate (FDR) procedure. In this study, we compared the FDR theories behind the statistical software of Statistical Analysis Software (SAS) and Golden Helix using a database obtained from a methadone maintenance treatment cohort in Taiwan. The association analyses were performed between the delta opioid receptor (OPRD1) of seven SNPs from rs2236861 to rs760588, or four SNPs from rs2236861 to rs419335, and the diastolic blood pressure (DBP) in both genotypes and allele types. In general linear model (GLM) of association analyses, SNP rs797397 had the most significant association with DBP among seven or four SNPs of both genotype (p=0.0012) and allele type (p=0.0004) using SAS and Golden Helix programs. The genotype and allele type of this SNP had 0.0039 and 0.0018 FDR, respectively after Benjamini and Hochberg (BH) procedure analyses using SAS program, and had 0.0083 and 0.0027 FDR, respectively after the positive false discovery rate (pFDR) analyses using Golden Helix program in seven SNPs. The genotype and allele type of this SNP had 0.0022 and 0.0015 FDR, respectively after BH analyses and 0.0047 and 0.0015 FDR, respectively after pFDR analyses in four SNPs. In summary, the pFDR had less significant level than BH procedure in the most significant GLM associations SNP in both genotypes and allele types at seven and four numbers of SNPs. The FDR values were the same in both BH and pFDR multiple corrections in the allele type of four SNPs. The BH procedure in SAS provided a more consistent FDR in both genotype and allele type of multiple correction analyses despite the number of SNPs.