Similar protocols

Protocol publication

[…] We focus on five established type II diabetes (T2D) loci (IGF2BP2, CDKN2A/B, KCNQ1, FTO, CDKAL1). These loci were explicitly chosen as they have association signals present in multiple ethnic groups, and several of these loci have shown large differences in LD between ethnic groups (eg, CDKAL1 and KCNQ1). Such loci characteristics are likely to favour success for trans-ethnic meta-analyses.In this simulation study, trans-ethnic meta-analysis was carried out by a frequentist fixed-effects meta-analysis that was implemented in GWAMA as a proof of principle in comparing fine-mapping of studies having varying degrees of ancestral diversity. In addition, for completeness, we also employed GWAMA using random effects in a selection of settings. Fine-mapping assessment was examined when the variant had a non-null effect in all of the populations and under various levels of ancestral diversity: European-only samples; moderate ancestral diversity (European and Asian samples); and high ancestral diversity (European, East Asian, and African samples). [...] Sets of six cohorts were meta-analysed across five loci under various allelic heterogeneity models and ancestry compositions of the contributing cohorts. Data were simulated using Hapgen2 from 6 populations based on the 1000 Genomes June 2011 haplotypes: CEU, TSI (European reference panels (RPs)), CHB, JPT (East Asian RPs), LWK, and YRI (African RPs). Each cohort was composed of 1000 cases/1000 controls, and 1000 replications were used for each setting. These simulations represent directly typed data, and variants are referred to as perfect. They were then thinned down to GWAS density based on the SNPs in the Illumina 660-Quad array, and subsequently imputed via IMPUTE2, using the same cross-population 1000 Genomes reference panel, and effective population size Ne=20 000. A 500 kb up- and down-stream buffer was included in the imputation, and variants with SNPTEST proper information score below 0.4 were filtered out. Both perfect (all variants are directly typed) and imputed data were analysed. Analysis of the perfect data illustrates an optimal scenario and provides the maximum possible power and refined fine-mapping resolution that may be attained, whereas the imputed data represent a more realistic setting.There were three general ancestry combinations considered for meta-analysis: Single ancestry (European): 6 CEU samples;Moderate ancestral diversity (European and Asian): 3 CEU+TSI samples, 3 CHB+JPT samples; andHigh ancestral diversity (European, Asian, and African): 1 sample from each of CEU, TSI, CHB, JPT, LWK, and YRI.Single ancestry (European): 6 CEU samples;Moderate ancestral diversity (European and Asian): 3 CEU+TSI samples, 3 CHB+JPT samples; andHigh ancestral diversity (European, Asian, and African): 1 sample from each of CEU, TSI, CHB, JPT, LWK, and YRI.Within each locus, causal variants were selected within ±0.5% of one of three minor allele frequencies (MAFs): 5, 10, or 20%, with respective relative risks of 1.4, 1.3, and 1.2. As the MAF varies between the populations, when there was a shared causal variant c1, it was selected to satisfy the MAF requirements in the CEU population frequency and to be nonmonomorphic in the other populations. A single causal variant c1 was considered for each of the single ancestry and moderate and high ancestral diversity scenarios. […]

Pipeline specifications

Software tools GWAMA, HAPGEN, IMPUTE, SNPTEST
Diseases Diabetes Mellitus, Type 2