Academia Sinica Logo

Based on an Ashkenazy Jewish case-control cohort for schizophrenia, we carried out (1) genetic association analysis for one variant at a time (GWAS) and (2) digenic analysis by comparing frequencies of genotype pairs between cases and controls. To control for genetic heterogeneity between sexes, we analyzed males and females separately. After pruning of variants in each of males and females, single-variant allelic analysis furnished 9 and 8 statistically significant variants in males and females, respectively, with 3 of these variants being significant in both males and females. Of the 14 distinct variants in males and females, 4 (29%) reside in genes. In digenic analysis, we found 76 significant genotype pairs, comprising 36 distinct variants, 20 (56%) of which reside in genes, many of which are known risk genes, thus lending credence to our approach. 

These analyses focused on statistically significant variants and genotype pairs. However, as has been pointed out close to ten years ago [1], a better approach would be to prioritize variants with large positive predictive values (PPV) even though such variants are generally not significant. This recommendation has largely been ignored in human gene mapping, which I will briefly discuss.