ESTIMATION OF GENETIC PARAMETERS FOR NUMBER OF PIGLETS BORN ALIVE USING RANDOM REGRESSION MODEL

LUKOVIĆ Z.¹, ², GORJANC G.¹, MALOVRH Š.¹, KOVAČ M.¹

¹University of Ljubljana, Biotechnical Faculty, Ljubljana, Slovenia

²University of Zagreb, Faculty of Agriculture, Zagreb, Croatia

Keywords: random regression model, genetic parameters, litter size, pigs

INTRODUCTION

In conventional animal breeding, both repeatability and multitrait(MT) models have been applied to analyze traits that are measured more than once in lifetime. More appropriate way of dealing with these traits is to fit a set of random regression (RR) coefficients describing production over time for each animal to allow individual variation in the course of trajectory (1). In pigs, RR modelswere mainly used for growth and feed intake. Although litter size differs from fattening traits, some authors suggest that RR models could be appliedin selection on this trait (2, 3). Possible genetic differences in litter size along the trajectory could be identified using RR models andsomething like persistency in litter size could be used to select breeding sows (3).

The aim of this paper was to estimate and compare dispersion parameters from MT andRR models with Legendre (LG) polynomials of different order for number of piglets born alive (NBA).

MATERIAL AND METHODS

Data from 98,359litters from first to sixth parity, collected between years 1990 and 2002 at farm Nemščak, wereanalyzed. The pedigree file contained 28,055 sows with records and 6,733 ancestors. Fixed part of the models was developed with the SASpackage (4). In MT, model 1 was used for gilts and model 2 for sows after firstparity. Model 3 was used in RR analysis for sows together.

/ (1)
/ (2)
/ (3)

where is NBA, ,, and are fixed effects of breed,mating season, parity and weaning to conception interval, respectively. For MT models age at farrowing () was adjusted by quadratic regression for each trait separately, whileit was nested within parity in RR model. Previous lactation length () was fitted as linear regression. Random part of the MT modelsconsisted of direct additive genetic ()and common litter effect (). RRs described byLG polynomialsof standardized parity () were included for direct additive genetic, permanent environmentaland common litter effect. LG polynomials from the first (LG(1)) to third order (LG(3)) were fitted. Estimation of (co)variance components fromMT and RR models was based on REML method using the VCE5 software package (5). Modul SAS/IML was used forcomputation of eigenvalues for covariance matrices of regression coefficients to quantify contribution of higher order of LG polynomials.

RESULTS AND DISCUSSION

Estimated phenotypic variancesand ratios were approximately the same between models.Eigenvalues of genetic covariance function show that the constant term accounted for around 93-95% of total genetic variability for NBA. This means that5-7% of variability in our study is covered by different production curves of sows.Computing time needed for MT analyze wasabout five times longer than for RR model with LG3.

Table1. Estimated variance and ratios in multitrait and random regression models

Multitrait model / Random regression with LG3
Parity / Var(ph) / h² / l² / Var(ph) / h² / l² / p²
1 / 7.85 / 0.102 / 0.021 / 7.87 / 0.101 / 0.021 / 0.054
2 / 7.38 / 0.130 / 0.030 / 7.39 / 0.125 / 0.013 / 0.070
3 / 7.99 / 0.112 / 0.007 / 8.03 / 0.123 / 0.005 / 0.079
4 / 8.19 / 0.111 / 0.026 / 8.20 / 0.118 / 0.005 / 0.091
5 / 8.25 / 0.123 / 0.039 / 8.27 / 0.118 / 0.010 / 0.099
6 / 8.54 / 0.118 / 0.034 / 8.56 / 0.119 / 0.020 / 0.099

Table 2. Eigenvalues of estimated covariance matrices of RR coefficients as proportion (%) in the total variability for random effects

Direct Additive / Perm. Environmental / Common litter
Eigenvalues / LG(1) / LG(2) / LG(3) / LG(1) / LG(2) / LG(3) / LG(1) / LG(2) / LG(3)
0th / 95.04 / 94.61 / 93.60 / 91.45 / 91.13 / 90.85 / 65.80 / 63.25 / 59.71
1st / 4.96 / 4.45 / 5.08 / 8.55 / 8.86 / 9.14 / 34.20 / 36.75 / 34.40
2nd / 0.93 / 1.31 / 0.01 / 0.01 / 0.00 / 5.88
3rd / 0.01 / 0.00 / 0.01

REFERENCES

1. Meyer, K., 1998. Modeling ‘repeated’ records: covariance functions and random regression models to analyse animal breeding data. In: Proc. 6th WCGALP 25:517-520, Armidale, Australia.

2. Schaeffer, L.R., 1994. Application of Random Regression Models in Animal Breeding. (25.10.2002.)

3. Huisman, A.E., 2002. Genetic analysis of growth and feed intake patterns in pigs. Ph.D. Thesis, Wageningen Institute of Animal Sciences, The Netherlands.

4. SAS Inst. Inc. (2001). The SAS System for Windows, Release 8.02. Cary, NC.

5. Kovač, M., Groeneveld, E. (2002). VCE-5 User’s Guide and Reference Manual Version 5.1. Institute of Animal Science and Animal Husbandry, FAL, D-31535 Neustadt, Germany.