Testing for association between type 2 diabetes and 225 candidate genes in 2357 Finnish cases and controls

K.J. Gaulton(1), K.N. Conneely(2), Y. Li(2), A.U. Jackson(2), L.J. Scott(2), W.L. Duren(2), P.S. Chines(3), N. Narisu(3), L. Bonnycastle(3), A. Swift(3), T.T. Valle(4), J. Tuomilehto(4), R.N. Bergman(5), F.S. Collins(3), M. Boehnke(2), K.L. Mohlke(1)

(1) U North Carolina, Chapel Hill, NC; (2) U Michigan, Ann Arbor, MI; (3) NHGRI, Bethesda, MD; (4) Nat'l Public Health Inst., Helsinki, Finland; (5) U Southern California, Los Angeles, CA.

Type 2 diabetes (T2D) is a complex disorder with a strong genetic component, although the underlying genes are incompletely known. As stage 1 of a 2-stage T2D association study (which includes a genome-wide association (GWA) study, see abstract from Scott et al.), we are genotyping 3378 SNPs from 225 candidate genes in 1171 T2D cases and 1186 normal glucose tolerant controls from the Finland-United States Investigation of NIDDM Genetics (FUSION) study. Genes were selected using CAESAR, our algorithm that uses text and data mining to integrate annotations from 11 databases and rank human genes based on their potential function. To test the ability of CAESAR to successfully select complex trait genes, we used the algorithm to rank genes for 11 complex traits. Of 17 tested genes previously implicated in these traits, 8 were present in the top 2% of ranked genes. We used CAESAR to select 209 candidate genes for T2D and genotyped SNPs in these and 16 other genes based on prior evidence of association in FUSION. 1850 SNPs within 10 kb upstream and 5 kb downstream of these genes are present on the GWA panel. 1528 additional SNPs were selected to more completely cover variation in these genes based on linkage disequilibrium (r2 >.8) and annotation. The 3378 SNPs tag an additional 7099 SNPs using the same r2 threshold. Preliminary results using the 1850 GWA SNPs in 885 cases and 885 controls indicate that the most significant SNP lies in the ARID2 gene and has an unadjusted p-value of 1.7x10-4 under a multiplicative model. Within each gene, we adjusted for the number of SNPs and the correlation between them and found that 3 genes exhibit p-values<.005, which is consistent with the 1.1 gene expected by chance (p=.1). Our completed stage 1 study will include all SNPs and 2357 samples to further determine the significance of these genes in T2D.