Eagle2, SHAPEIT212, SHAPEIT2 with its –no–mcmc option (which increases speed at the expense of accuracy), and a reference-based version of Eagle1 that we implemented for comparison. We also attempted to benchmark Beagle v4.125 but found it was too slow for this benchmark to be practical: for the smallest analysis (chromosome 20 with Nref = 15,000 and Ntarget = 72), Beagle v4.1 required 3.6 days (in contrast to 1.1 minutes for Eagle2). (We note that the focus of Beagle v4.125 is its haploid imputation algorithm, which is much faster than its phasing algorithm.) We did not benchmark HAPI-UR11 as HAPI-UR does not implement reference-based phasing.
