Conditional association, in which the association between SNP and trait is evaluated after conditioning on the top SNP at a locus, can be used to identify multiple signals of association at a previously identified GWAS locus. Conditional association methods have traditionally required individual-level data in order to jointly fit multiple SNPs. Recent work has shown that conditional and joint association analysis of multiple SNPs can be approximated using only summary association statistics together with linkage disequilibrium (LD) information estimated from a population reference panel such as 1000 Genomes (see Box 1)7. This has enabled the discovery of new secondary associations at known loci for height, BMI, and other complex traits and diseases, increasing the variance explained by GWAS associations for these traits8–10; for example, in a recent height GWAS, approximate conditional analysis using summary data identified 697 genome-wide significant associations, including 34 secondary associations with r2>0.1 to a more significant SNP at the same locus8.
