Although none of the discovery markers exceeded the GWS threshold, we wanted to further test the existence of true positive signal in our top associated regions and sought replication in independent samples. Previously, when comparing the results of an early iteration of the PGC schizophrenia GWAS (PGC1) [40], which is of similar size to that of reported here for the PGC ASD GWAS, to more recent and larger PGC schizophrenia GWAS (PGC2) [19], we noted that of the association signals 5 × 10−8 ≤ P < 10−4 in the schizophrenia PGC1 study, over 10% (20 of 183 independent loci) were subsequently reported as GWS in the schizophrenia PGC2 analysis. Others have also observed this, revealing in their sample that a substantial minority of associations with borderline genome-wide significance represent replicable and possibly genuine associations [41].
