Cyclic genotyping strategies. II: True and perceived utilities under incorrect allele frequency assumptions

Abstract

This paper investigates the potential problems associated with assuming incorrect population frequencies for segregation analysis when animals are genotyped one by one in a cyclic fashion with segregation analysis carried out at each cycle. The base population allele frequencies of 0.1 and 0.5 studied, with the incorrect frequencies assumed for segregation analysis of 0.5 and 0.1 respectively, were investigated on the basis of their covering the range of possible frequencies and the most erroneous assumption of frequencies. An index modelled using linear regression (LR) was employed to choose the next animal to be genotyped in each cycle, based on segregation analysis at the incorrect frequency. The resultant utility was evaluated both at the correct frequency giving actual utility and at the incorrect frequency, the perceived utility. The results are compared with those in which all segregation analysis was carried out at the correct base population allele frequency. The assumption of incorrect population frequency for segregation analysis leads to a decline in the predictive performances of both indices in terms of the true utility of prediction. When utility is also computed at the incorrect frequency to give perceived utility, the assumption of incorrect population frequency of q = 0.5 for segregation analysis of populations simulated at q = 0.1 leads to a perceived decline in the predictive performance of the LR index. However, the assumption of incorrect population frequency of q = 0.1 for segregation analysis of populations simulated at q = 0.5 leads to a perceived increased utility. This latter paradox is explained in terms of genotype probabilities with reference to the method of construction of the genotype probability index.