Predicting body protein and body fat for breeding sows of a modern commercial genotype

Modern sow genotypes have changed considerably since body composition studies were conducted over 25 years ago on Australian genotypes. Earlier predictive equations were developed on datasets of slaughtered gilts and first-litter sows (e.g. Mullan and Williams 1990) or used genotypes from the United Kingdom (Large White × Landrace crossed with Landrace × Meishan; Gill 2006). As part of a larger project to evaluate sow lifetime performance and longevity over several parities (Smits et al. 2017), we established predictive equations based on measured live animal data for body protein and fat of sows before mating for Australian maternal genotypes. The hypothesis was that predicted and actual tissues reserves were similar.

Over six replicates, six unmated gilts (40 weeks of age) and 52 mixed parity (1 to 9) sows at weaning were selected for the study (Large White × Landrace F1 cross, PrimeGro Genetics™). Prior to slaughter, animals were measured for several traits including ultrasound (PieMedical, linear array probe 5 MHz) fat depth thickness measured at 65 mm from midline at last rib (P2) and 20 mm from midline at junction of the tail (LEG); loin muscle depth at P2; shoulder height; girth circumference at foreleg and last rib; live weight (LWT); and parity. After 24 h, carcasses were split into primals and measured for protein and fat content using dual X-ray absorptiometry (Suster et al. 2003). Viscera were collected from the abattoir and weighed full and empty, frozen and then ground, freeze-dried and analysed for protein and fat using chemical analysis methods. Live empty bodyweight was calculated from the empty cold carcass weight plus the estimated blood volume (7% of live weight). Regression was used to develop the predictive equations with the highest regression (r²) for protein and fat content in the empty live weight. The least significant effect was removed from the model containing all factors in a step-wise fashion to develop a parsimonius model.

The predictive equations described with the highest regression in the model were as follows:

where α = –2.16 for parity 0; –0.83 for parities 1–2; 0.33 for parities 3–5; 0 for parities > 5

where α = 12.7 for parity 0; 4.51 for parities 1–2; 1.69 for parities 3–5; 0 for parities > 5.

There was a high consistency between actual protein and fat contents in the empty bodyweight and predicted values (Table 1). Adding additional parameters such as loin muscle diameter, girth dimensions and shoulder height did not increase the accuracy of the equation for predicting body tissue mass when live weight was recorded.

Table 1.  Comparison of actual tissue values (mean ± s.e.) in the empty bodyweight of unmated gilts and sows with predictive body protein (BPROT) and fat (BFAT) tissue reserves

These equations provide a valuable tool for predicting changes in body protein and fat reserves in this commercial genotype across a range of parities, weights and backfat measures.