Removing array-specific batch effects in GWAS mega-analyses by applying a two-step imputation workflow

Combining genetic data from different genotyping arrays (mega-analysis) increases statistical power but introduces array-specific batch effects that may bias results. This project developed a two-step genotype imputation workflow addressing this bias in studies using multiple genotyping platforms.Genotype data of 10 647 individuals generated using five different arrays were included. The two-step method involved creating intermediate array-type specific panels, which were then imputed against the 1000 Genomes reference panel. Batch effects were assessed using genetic principal component analysis of the combined imputed dataset. Performance was evaluated by comparing imputation quality and allele frequency differences between the two-step and the conventional array-specific imputation. Additionally, concordance with a whole-genome-sequenced subgroup was examined. Genome-wide association analysis on goiter risk and thyroid gland volume was conducted to compare outcomes between both imputation approaches.The workflow eliminated array-driven batch effect from the first 20 PCs and showed high correlation with the conventional approach for allele frequencies (r ² > 0.99). GWAS using the two-step imputation confirmed known associations on thyroid traits and revealed novel loci for thyroid volume (TG, PAX8, IGFBP5, NRG1), and goiter (XKR6), the latter not significant in the conventional imputation.

The study provides a workflow for high-quality imputation results without batch effects, fostering genetic analysis involving multiple genotyping arrays.