Estimating genetic merit when genotype data are incomplete

Abstract

Direct markers for QTL are becoming available and, for traits that are difficult to measure or observed later in life, offer the potential to increase the accuracy of EBVs. While genotyping costs are non-trivial, QTL information will not be available for all individuals - with important consequences for genetic evaluation systems. Using simulated populations in which all parents and some progeny are genotyped for a QTL acting additively, two methods of using genotypic information are compared with EBVs based on the infintessimal model. Both methods - deregressing EBVs for the QTL effect according to their accuracy, or using heterogeneous residual variances - improve correlations between simulated and estimated genetic merit, particularly for individuals without performance data.