Owing to the focus of our algorithm on the strongest association signal, an alternative approach to deal with multiple associations consists of using a stepwise regression strategy, which would then reveal the secondary association signals. Our colocalisation test can then be run on using the conditional p-values. We find this approach to be the most practical and illustrate below an application for a locus that contains several independent eQTL associations (Figure 6). In situations where only single SNP summary statistics are available, the approximate conditional meta-analysis framework proposed by Visscher et al. [23] can be used to obtain conditional p-values.
