Genotype error biases trio-based estimates of haplotype phase accuracy

Abstract

Haplotypes can be estimated from unphased genotype data using statistical methods. When parent-offspring data are available for inferring true phase from Mendelian inheritance rules, the accuracy of statistical phasing is usually measured by the switch error rate, which is the proportion of pairs of consecutive heterozygotes that are incorrectly phased. We present a method for estimating the genotype error rate from parent-offspring trios and a method for estimating the bias in the observed switch error rate that is caused by genotype error. We apply these methods to 485,301 genotyped UK Biobank samples that include 899 White British trios and to 38,387 sequenced TOPMed samples that include 217 African Caribbean trios and 669 European American trios. We show that genotype error inflates the observed switch error rate and that the relative bias increases with sample size. For the UK Biobank White British trios, we estimate that the observed switch error rate in the trio offspring is 2.4 times larger than the true switch error rate (1.41 x 10-3 vs 5.79 x 10-4) and that the average distance between phase errors is 64 megabases.