Impact of population genetic substructure on association studies and risk assessment for melanoma.

Abstract

8521 Background: Genetic substructure due to varying allele frequencies between populations can confound association studies. Ancestry informative genetic marker (AIMs) data combined with statistical adjustment can reveal spurious associations and identify population specific risk markers. In melanoma, AIMs may also be risk markers, e.g. pigment genes contribute to melanoma susceptibility and segregate with ancestry. We have thus developed a strategy to adjust for population genetic substructure (PGS) using AIMs, while identifying potentially novel genes associated with melanoma. Methods: 326 melanoma patients and 400 controls of European ancestry from the New York area were studied. Tag SNPs spanning 14 candidate genes and 75 AIMs were genotyped and odds ratios (OR), unadjusted and adjusted for PGS, computed. Results: A PGS model based on all AIMs separated cases and controls, suggesting that some AIMs were associated with melanoma. An algorithm was developed to select AIMs least capable of separating cases and controls to infer PGS and validated using simulations. The resulting model, which was reproduced using 49 additional AIMs, separated Northern (NE) and Southern Europeans (SE) and was used to adjust ORs. Three classes of SNPs were identified 1. Associated before and after PGS correction in both groups (10 SNPs localized to MATP, TYR and ERCC5). 2. Not associated in unadjusted analysis, but significantly associated with melanoma in NEs (6 SNPs localized to XPC, ERCC4, OCA2, ASIP and TYR). 3. Associated with melanoma before but not after adjustment. To determine if AIMs that separate cases and controls can identify novel melanoma genes, we genotyped 16 SNPs localized to 4 genes that house candidate AIMs. Four SNPs at 2 different loci were associated with melanoma (e.g. AIM1: OR=0.35, p=0.01; AIM2: OR=1.77, p=0.03). Conclusions: Our approach demonstrated that ancestry is a significant confounding factor in identifying melanoma susceptibility genes. Melanoma risk markers vary significantly between groups and a DNA based risk assessment model will require adjustment for ancestry. We have also identified potentially novel susceptibility melanoma genes for futher study.