Usefulness of running animal models in absence of pedigrees: Estimation of genetic parameters for gastrointestinal parasite resistance traits in Djallonke sheep of Burkina Faso

The main aim of the current study was to exemplify the usefulness of pedigree-free animal models in scenarios in which available information is limited. Up to 271 Djallonke sheep individuals of Burkina Faso were typed for 29 microsatellites. Performance records were obtained from a trial designed to assess environmental factors affecting gastrointestinal parasite resistance. Packed cell volume (PCV), log-transformed fecal egg count (lnFEC) and FAffa Malan CHArt (FAMACHA) eye scores were analyzed fitting univariate and multivariate animal models. Molecular information was used to: a) construct an artificial pedigree maximizing the correlation between the molecular coancestry matrix and the coancestry of the artificial genealogies (P* Matrix); and b) computing the Ritland's estimator of between-individuals molecular coancestry (R Matrix). All animal models fitted included as fixed effects the contemporary group, day of assessment after deworming and body weight as a linear covariate. As random effects, models included the additive genetic effect and the permanent environment associated to individual. Multivariate models further included the covariances between permanent environmental and additive genetic effects. Univariate models had difficulties to separate between additive genetic and permanent environmental effects. The P* Matrix gave higher estimates of additive genetic variance than the R Matrix. Independently of the Matrix used, multivariate models gave more consistent estimates of heritability (h(2)) and permanent environmental effect: estimates of h(2) for lnFEC varied from 0.063 +/- 0.037 to 0.173 +/- 0.076; estimates of h(2) for FAMACHA scores ranged from 0.206 +/- 0.070 to 0. 343 +/- 0.111; and c) estimates of h(2) for PCV ranged from 0.073 +/- 0.045 to 0.142 +/- 0.084. Genetic correlations estimated for the pair lnFEC-FAMACHA using multivariate models were significant in all cases varying from 0.548 +/- 0.247 to 0.785 +/- 0.242. Pedigree-free animal models may be advantageous in developing countries at the beginning of selection programmes or to improve the outcome of trials designed to assess performance of local breeds for economically interesting traits.