Validation of a Bayesian approach for maternity identification in abandoned lambs

Abstract

Abandoned lambs in sheep flocks use to involve few ewes at the same time; this phenomenon prevents the accurate collection of pedigree data and originates a moderate-to-low percentage of lambs with unknown or uncertain ancestors. Given the economic restrictions inherent to the sheep industry, the systematic implementation of laboratory paternity testing technologies to all abandoned lambs is not affordable by stockbreeders; if possible, only a reduced number of lambs can be validated by genotyping, and we need specific tools to elucidate the most relevant individuals (e.g. lambs with the highest chance to be offspring of high genetic merit ewes). We adapted a Bayesian mixed linear model to infer the dam of abandoned lambs by integrating both genetic and environmental sources of information from phenotypic data, and modelling the uncertain dam as an additional unknown parameter. Model performance was evaluated on simulated data and by assuming seven different scenarios where one to four abandoned lambs had to be assigned to two candidate ewes. The average probability of assignment to the correct dam (PACD) was 0.59, although within-scenario average PACD ranged between 0.51 and 0.70, and raw PACD estimates fluctuated between 0.04 and 1.0. Sensitivity varied across simulation scenarios, although most of the cases revealed values larger than 60%. This approach must be viewed as a useful tool for screening abandoned lambs and their candidate mothers, inferring the most plausible dam for each individual. Note that any inference on uncertain dams may reduce further paternity testing expenses.