Among Yale-Penn participants of European descent, several genes on chromosome 16 survived multiple testing correction for both association and interactive effects (Fig. 2): LCAT (passociation = 3.73 × 10−7; pinteraction = 1.82 × 10−7); TSNAXIP1 (passociation = 2.08 × 10−7; pinteraction = 2.13 × 10−7), CENPT (passociation = 2.39 × 10−7; pinteraction = 2.32 × 10−7), and PARD6A (passociation = 7.17 × 10−7; pinteraction = 5.57 × 10−7). The association of this gene cluster is driven by the effect of a single variant, rs8052287 (passociation = 2.15 × 10−7; pinteraction = 7 × 10−8; Fig. 3). Within this locus, 98% of the variance is explained by multivariate SD-gene interactions. We calculated the Bayes factors (BF) between the full model and models including the individual environmental exposures removed to explore which environmental variables are most relevant for the gene-environment signals of rs8052287. We observed putative gene-environment effects in rs8052287 (BF > 0) for OD criterion counts (BF = 12.2), CoD criterion counts (BF = 12.1), ND criterion counts (BF = 9.2), and co-occurrence of multiple SD diagnoses (BF = 2.1). We
