For cohorts with explicitly related samples we introduce a new hidden Markov model (which we call duoHMM) that can estimate the inheritance pattern between between the haplotypes of each parent-child duo. This method can be used to visualise the inheritance status across a chromosome, correct phasing errors that are inconsistent with pedigree information, and detect genotyping errors. We show that after applying this adjustment, SHAPEIT2's haplotypes are accurate enough that we can detect explicit recombination events between parent-child pairs. Applying this method to the SHAPEIT2 inferred haplotypes provides the most accurate performance in the extended pedigree setting. Using our method we are able to demonstrate that the recombination events that we infer from otherwise uninformative duos and trios can add power to association scans for recombination phenotypes. Specifically, at the established PRDM9 locus we are able to show that including these extra recombination events increases the signal of association for a hot spot usage phenotype. Overall, the combination of SHAPEIT2 and duoHMM provides a very general method for accurate phasing of cohorts with any levels of implicit or explicit relatedness between individuals.
