Genetic parameters of somatic cell scores using random regression test-day models with Legendre polynomials in Tunisian dairy cattle

•Genetic parameters of test-day somatic cell scores (TDSCS) in Tunisian dairy cattle were successfully estimated using random regression models (RRMs) with homogeneous residual variances.•Three models (LP3, LP4 and LP5) where compared based on the log marginal density for Bayes factor, the deviance information criterion, and the similarities between the estimated breeding values and genetic trends.•The somatic cell scores are much more emphasized by temporary environmental factors at the beginning of lactation than during any other lactation period.•The less parametrized model (LP3), with low computational demand, provided similar estimates of the genetic parameters as LP5 model, and was chosen as the parsimonious model.•The RRM study provided a good insight into the changes in genetic merit of the somatic cell score trait and that an EBV for the SCS should be included in the selection programs to perform national genetic evaluations of dairy cattle. The aim of this study was to analyse the genetic parameters of test-day somatic cell scores (TDSCS) in Tunisian dairy cattle using random regression models (RRM) with homogeneous residual variances. Data included 43,647 test-day somatic cell count (TDSCC) records collected on 4825 Holstein-Friesian cows with parities up to the fifth, and calving dates between 2000 and 2014. Records of the same animal belonging to different lactations were treated as repeated records. Additive genetic (AG) and permanent environmental (PE) random effects were fitted using Legendre polynomials functions (LP) up to the fifth-order. The models included fixed regressions on days-in-milk (DIM), modelling the average phenotypic curve using Ali-Schaeffer's lactation function. Overall, three models (LP3, LP4 and LP5) where compared using the -2log marginal density for Bayes factor (-2logp) and the deviance information criterion (DIC). Models were also compared by similarities between the estimated breeding values and genetic trends. For all models, the total variances tended to decrease during the first month of the lactation, then they maintained a nearly constant trend. All models showed the largest AG and PE variances at the beginning and at the end of lactation with an oscillating pattern in the middle. The heritability estimates of from the three models ranged from 0.01 to 0.08 with the lowest heritability estimates being in the mid-lactation period. Genetic correlations between adjacent DIM were high and then decreased as the inte