For our reference-based phasing benchmarks using genotyped UK Biobank data, we constructed reference panels by randomly selecting Nref = 15,000, 30,000, 50,000, or 100,000 samples (disjoint from the 72 UK Biobank trios) and phasing these samples using Eagle1 (as phasing using SHAPEIT2 would have required several weeks13). We then applied each reference-based phasing method to phase chromosomes 1, 5, 10, 15, and 20 of the Ntarget = 72 trio children using each simulated reference panel, and we compared the phased output against trio phase calls to compute switch error rates2,13. In our results, we report mean switch error rates and s.e.m. over the 70 European-ancestry trio children (according to self-reported ethnicity; see above).
