Additional GWS findings implicate other variants that map to the ADH cluster and extend to a region centromeric to it that includes non-ADH genes. Variants mapped to PDLIM5, spanning 32 kb, mark the centromeric extent of these associations. This locus has previously been reported to be associated with schizophrenia26,27. We identified significant association in the EA sample, under the adjusted ordinal model, for a series of SNPs at this locus, a result driven by the SAGE data set. These associations were not identified in the original SAGE reports,28 presumably because they used diagnostic models focusing primarily on case–control analysis. Associations were observed across the LOC100507053 locus, most from the AA part of the sample, with a P-value as low as 1.06 × 10−12 at rs28864441, but one SNP was associated only in EAs (P = 2.46 × 10−9). The support for association at this locus in multiple samples and two major populations, AAs and EAs, appears strong, but is apparently accounted for by variation elsewhere in the ADH cluster, that is, it most likely reflects linkage disequilibrium with other variants.
