A random regression model for total litter weight in mice

Measuring litter weight (LW) as a summary trait in multiparous animal species is less cost-intensive than individual birth weight. We propose a random regression model for total LW, where the genetic effects of dams and sires are modeled as linear random regression slopes on litter size with potentially different variances. Restricted maximum likelihood estimation of variance components was conducted separately for four diverse mouse lines, and the results were compared. In total, there were 11,430 total LWs, representing 175,446 pups. The sire genetic variance component was significant in two lines. The random regression model implies that the heritability of the trait and the reliability of estimated genetic effects depend on litter size, in contrast to previously used models. When averaged over different litter sizes, heritability for total LW was moderate to low at 0.32 in the unselected control line FZTDU, at 0.19 in the growth-selected line DU6, and at 0.14 and 0.12 in the high-fecundity lines DUC and DUK, respectively. In conclusion, the proposed random regression model makes more detailed use of the information available in total LW data. It provides potential applications in other multiparous species, for a cost-effective selection for birth weight using LW data.