One major use of our new method (and of imputation methods generally) will be to facilitate meta-analyses [9],[10], which combine samples from studies of similar diseases to increase the chances of detecting low-penetrance risk alleles. For this application, we might expect to repeat Scenario B for a number of different study samples genotyped on different SNP chips. Rather than re-phase the diploid reference panel for each study sample, we can save time by simply storing the posterior samples from a single run of phasing the reference panel, then read these sampled haplotypes from memory when processing each study sample. This functionality is implemented in our software.
