Supplementary Material for: A Case Study of Fixed-Effects and Random-Effects Meta-Analysis Models for Genome-Wide Association Studies in Celiac Disease
datasetposted on 06.10.2015 by Ahn R.S., Garner C.
Datasets usually provide raw data for analysis. This raw data often comes in spreadsheet form, but can be any collection of data, on which analysis can be performed.
Background/Aims: Amongst the many approaches to genome-wide association study (GWAS) meta-analysis (MA), the most popular methods are based on fixed-effects (FE) modeling because it tends to be the statistically most powerful approach in the absence of heterogeneity. However, FE-based MA ignores the potential heterogeneity that may exist between studies. The purpose of our analysis was to test whether results from random effects (RE)-based methods that account for heterogeneity differed significantly from the results that were originally published. Methods: We reanalyzed two GWAS FE-based MAs of celiac disease with RE-based methods: (1) a two-stage GWAS MA that includes 9,451 celiac disease cases and 16,434 controls from 12 collections and (2) a single-stage GWAS MA using a custom dense genotyping platform to capture low-frequency and rare variants in 12,041 cases and 12,228 controls from 7 collections. Results: We present evidence that SNPs at loci that were previously reported to be genome-wide significant (GWS; p < 5 × 10-8) in either the two-stage GWAS MA or the single-stage GWAS MA were not GWS when heterogeneity was accounted for by an RE MA method. Conclusion: This case study highlights the strengths of RE MA methods in the presence of heterogeneity and of pooled FE methods.