We have presented a method of approximate conditional and joint genome-wide association analysis that is powerful, versatile and computationally fast. The method does not require any additional genotyping or phenotyping and does not rely on individual-level genotype and phenotype data, except for a reference population with individual genotypes—either from one of the participating studies of the meta-analysis or from genotype data in the public domain—that is required for LD estimation. The effect sizes of most SNPs that are associated with complex traits are very small, such that there is a great benefit in using estimates from a large-scale meta-analysis, and a reasonably large reference sample is sufficient to estimate LD between SNPs located near to one another. We believe that this method is useful to refine independent GWAS associations and to identify additional associated variants in large-scale meta-analysis where the pooled individual-level genotype data are unavailable for analysis.
