347. Genotype error correction and the accurate identification of recombination sites

Abstract

Measuring recombination rate and identifying hotspots requires reliable crossover detection, which is dependent on accurate phasing and genotyping. Given the low number of crossover events that occur at meiosis even low levels of genotyping and phasing errors can frustrate detection. We outline a pipeline for phasing and crossover detection and evaluate the performance across simulated populations, which contain from 0 to 9% genotyping errors. We present a novel algorithm for genotype correction in pedigreed populations, genoFix, which combines a Bayesian network estimation of Mendelian errors with an empirical LD approach. The algorithm typically achieved a precision of >90% and recall of >80% at 1% error for a chicken population with broiler-type haplotype diversity and was still effective even at atypically high rates of diversity, using populations simulated from randomized founders. Correction of genotyping errors improved both phasing accuracy and crossover detection.