Identification of Genetic Polymorphisms Contributing to Risk in MDS Using Innovative GWAS Approaches.

Abstract 734

Genetic predisposition to MDS and AML is likely polygenic and may involve several low penetrance alleles which in concert with exogenous factors result in highly variable phenotype and late presentation, not easily amenable to genetic studies. With the advent of whole genome scanning (WGS) technologies utilizing various SNP array (SNP-A) platforms, large scale investigations in various disorders have been conducted. A systems level understanding of particular disease allows for identification of candidate genetic variants as prognostic or diagnostic markers. We have applied 6.0 SNP-A containing 924.644 SNP probes to conduct a comprehensive genome-wide association study (GWAS) in MDS including sAML with the aim of identifying low prevalence genetic variants that contribute to individual disease risk. We have studied 189 patients with MDS and sAML and a cohort of 2230 controls. After exclusion of SNP's with call rate of <95% and those with violation of Hardy Weinberg equilibrium (p<.01), 809.802 SNPs (87.5% of initial set) were passed for further investigation. Single allele χ² for all autosomal markers was performed. A set of 3600 SNP's pointing towards genes with minor allele frequency (MAF) <10% and p<.001 after Bonferroni correction (more conservative multiple hypothesis correction than False Discovery Rate) were selected. Out of 38 significant non-synonymous SNPs 3 were selected, while 3/64 exonic non synonymous SNPs were prioritized for final investigation. These included rs805267, rs2499953 and rs2681417 pointing towards LY6G5B with (OR 4.9), MMP26 (OR 4.7) and CD86 (OR2.9), respectively. LY6G5B gene was represented by marker rs4656334 (p<1×10^–12) occurring at the heterozygous frequency of 16.3% vs. 5.4% in controls, and in homozygous frequency of 4.1% vs. 0.04% in controls resulting in MAF of 12.2% vs. 2.7% in controls (p<1×10^-12). Rs2499953 and rs2499956 pointed towards MMP26 and was found in the heterozygous variant in 17.0% vs. 3.7% in controls. CD86 gene was represented by singular rs2681417 marker (p<1×10^–7) occurring at the heterozygous frequency of 29.0% vs. 13.6% frequency in controls, and in homozygous frequency of 4.0% vs. 0.4% in controls. The 3/3536 strongest intronic SNPs included rs4656334, rs4647493 and rs700060 and directed via LD to informative genes ATF6 with odds 3.15, FANCC (OR 135.3) and RABGAP1 (OR 37.1) respectively. Of interest is that ATF6/αaRheb-mTOR signalling promotes survival of dormant tumour cells in vivo. ATF6 gene was represented by marker rs4656334 (p<1×10^–7) and occurred at a heterozygous frequency of 7% vs. 8.3% in controls, while the homozygous constellation was 7 × higher in patients (14.6% vs. 2.4% in controls) with the corresponding MAF of 18.0 vs. 6.5%. Another 2 SNPs within this locus were highly significant (rs16860777, p<1×10^–6; and rs12401299, p<1×10^–6). Second potential locus identified in our study (FANCC gene) was represented by singular rs4647493 marker (p<1×10^–20) occurring at the heterozygous frequency of 16.9% vs. 0.05% in controls. RABGAP1 gene was represented by singular rs700060 marker (p<1×10^–17) occurring at the heterozygous frequency of 13.9% vs. allelic frequency of 0.3% in controls. While these results of analysis targeting individual SNP provide intriguing research avenues, such an approach offers only limited understanding of the complex genetic traits as not an individual SNP, but rather a joint action of several SNPs results in particular outcomes. Consequently, in study of MDS, we applied the network gene association analysis as a new paradigm incorporating both “operator OR” and “operator AND” thereby allowing for dependence and independence testing and possibly, to identification of meaningful pathways. We performed a simulation study, where genotypes were drawn including homozygous reference, heterozygous and homozygous variant for each SNP S_i = 1,… 64 where the MAF of SNP is chosen uniformly at random. We have identified rs236113 and rs2499953 (MCM8 and MMP26), both in homozygous variant with occurrence of 19% in patients and 1.8% in controls at a specificity score 98.2% and p<1×10^–32. Accordingly the presence of both SNP's increases relative risk given the specificity when both SNPs are present. In sum, our study constituted the first network analyses of predisposing factors taken in consideration as groups and identified informative loci that can lead to delineation of causative genetic profiles.