accurate by a large margin in the African panels (YRI, LWK, and MKK). These trends cannot be attributed to the fact that we are running Beagle with a stratified reference panel when the method is not designed for that situation: IMPUTE2 also produced higher accuracy when we used reference panels that were well-matched to the target panels, both in the current HapMap 3 framework (data not shown) and in our MalariaGEN analyses (results below). We further note that the Beagle results shown here are better than the ones we obtained with smaller, less diverse HapMap 3 reference sets (data not shown). We tried running Beagle with larger values of its niterations and nsamples parameters, but there was essentially no change in these results (data not shown). We speculate on the mechanistic reasons for the accuracy differences between IMPUTE2 and Beagle in the Discussion.
