p-values in the MH test do not systematically correspond to high rankings in the individual tests. The p-values in the dTDT share the highest variance (2.01×10− 3) among the three tests because of the randomness introduced by the inference procedure as well as the difference in sample sizes across the tests. A larger sample size will likely increase the power and generate more robust test results. In total, we have 257 SNPs in 75 genes with p-values <10− 4. 14 SNPs at the same level of p-values are found in replication of Edenberg et al’s study [79]. Four of these 14 SNPs have associated genes: CSMD2 on chromosome 1, LZTS2 & PDZD7 on chromosome 10, and Gcom1 on chromosome 15. There are 34 vs. 11 SNPs that have p-values <10− 5. Five SNPs across chromosome 1, 3, 9, 12 and 14 show p-values <5.1×10− 7 which is more statistically significant than the case-control analysis by Edenberg et al [79]. Our results also show clusters of SNPs by distance with p-values <10− 5 (more than five such markers in one cluster) in genes EXOC6B, FTO, NCAM2 and PPEF1 on chromosome 2, 5, 21 and X, respectively.
