Search

In broiler breeding, superior individuals for growth become parents and are later evaluated for reproduction in an independent evaluation; however, ignoring broiler data can produce inaccurate and biased predictions. This research aimed to determine the most accurate, unbiased, and time-efficient approach for jointly evaluating reproductive and broiler traits. The data comprised a pedigree with 577K birds, 146K genotypes, phenotypes for three reproductive (egg production [EP], fertility [FE], hatch of fertile eggs [HF]; 9K each), and four broiler traits (body weight [BW], breast meat percent [BP], fat percent [FP], residual feed intake [RF]; up to 467K). Broiler data were added sequentially to assess the impact on the quality of predictions for reproductive traits. The baseline scenario (RE) included pedigrees, genotypes, and phenotypes for reproductive traits of selected animals; in RE2, we added their broiler phenotypes; in RE_BR, broiler phenotypes of nonselected animals, and in RE_BR_GE, their genotypes. We computed accuracy, bias, and dispersion of predictions for hens from the last two breeding cycles and their sires. We tested three core definitions for the algorithm of proven and young to find the most time-efficient approach: two random cores with 7K and 12K animals and one with 19K animals, containing parents and young animals. From RE to RE_BR_GE, changes in accuracy were null or minimal for EP (0.51 in hens, 0.59 in roosters) and HF (0.47 in hens, 0.49 in roosters); for FE in hens (roosters), it changed from 0.4 (0.49) to 0.47 (0.53). In hens (roosters), bias (additive SD units) decreased from 0.69 (0.7) to 0.04 (0.05) for EP, 1.48 (1.44) to 0.11 (0.03) for FE, and 1.06 (0.96) to 0.09 (0.02) for HF. Dispersion remained stable in hens (roosters) at ~0.93 (~1.03) for EP, and it improved from 0.57 (0.72) to 0.87 (1.0) for FE and from 0.8 (0.79) to 0.88 (0.87) for HF. Ignoring broiler data deteriorated the predictions' quality. The impact was significant for the low heritability trait (0.02; FE); bias (up to 1.5) and dispersion (as low as 0.57) were farther from the ideal value, and accuracy losses were up to 17.5%. Accuracy was maintained in traits with moderate heritability (~0.3; EP and HF), and bias and dispersion were less substantial. Adding information from the broiler phase maximized accuracy and unbiased predictions. The most time-efficient approach is a random core with 7K animals in the algorithm for proven and young., (© The Author(s) 2023. Published by Oxford University Press on behalf of the American Society of Animal Science.)

Computer vision systems (CVS) have been shown to be a powerful tool for the measurement of live pig body weight (BW) with no animal stress. With advances in precision farming, it is now possible to evaluate the growth performance of individual pigs more accurately. However, important traits such as muscle and fat deposition can still be evaluated only via ultrasound, computed tomography, or dual-energy x-ray absorptiometry. Therefore, the objectives of this study were: 1) to develop a CVS for prediction of live BW, muscle depth (MD), and back fat (BF) from top view 3D images of finishing pigs and 2) to compare the predictive ability of different approaches, such as traditional multiple linear regression, partial least squares, and machine learning techniques, including elastic networks, artificial neural networks, and deep learning (DL). A dataset containing over 12,000 images from 557 finishing pigs (average BW of 120 ± 12 kg) was split into training and testing sets using a 5-fold cross-validation (CV) technique so that 80% and 20% of the dataset were used for training and testing in each fold. Several image features, such as volume, area, length, widths, heights, polar image descriptors, and polar Fourier transforms, were extracted from the images and used as predictor variables in the different approaches evaluated. In addition, DL image encoders that take raw 3D images as input were also tested. This latter method achieved the best overall performance, with the lowest mean absolute scaled error (MASE) and root mean square error for all traits, and the highest predictive squared correlation (R2). The median predicted MASE achieved by this method was 2.69, 5.02, and 13.56, and R2 of 0.86, 0.50, and 0.45, for BW, MD, and BF, respectively. In conclusion, it was demonstrated that it is possible to successfully predict BW, MD, and BF via CVS on a fully automated setting using 3D images collected in farm conditions. Moreover, DL algorithms simplified and optimized the data analytics workflow, with raw 3D images used as direct inputs, without requiring prior image processing., (© The Author(s) 2020. Published by Oxford University Press on behalf of the American Society of Animal Science. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com.)

Computer vision applications in livestock are appealing since they enable measurement of traits of interest without the need to directly interact with the animals. This allows the possibility of multiple measurements of traits of interest with minimal animal stress. In the current study, an automated computer vision system was devised and evaluated for extraction of features of interest, as body measurements and shape descriptors, and prediction of body weight in pigs. From the 655 pigs that had data collected 580 had more than 5 frames recorded and were used for development of the predictive models. The cross-validation for the models developed with data from nursery and finishing pigs presented an R2 ranging from 0.86 (random selected image) to 0.94 (median of images truncated on the third quartile), whereas with the dataset without nursery pigs, the R2 estimates ranged from 0.70 (random selected image) to 0.84 (median of images truncated on the third quartile). However, overall the mean absolute error was lower for the models fitted without data on nursery animals. From the body measures extracted from the image, body volume, area, and length were the most informative for prediction of body weight. The inclusion of the remaining body measurements (width and heights) or shape descriptors to the model promoted significant improvement of the predictions, whereas the further inclusion of sex and line effects were not significant.

The objective of this study was to quantify the effect of heat stress during the life of a pig on its final weight, as a first step toward a genetic evaluation for heat tolerance. Data included carcass weights of 23,556 crossbred pigs [Duroc x (Landrace x Large White)] raised on 2 farms in North Carolina and slaughtered from May 2005 through December 2006. Weather data were available from a nearby weather station. Lifetime of a pig was assumed to be partitioned into 2 periods. During an initial period, the effect of heat stress was assumed to be negligible or compensated for later. During the second period ending in slaughtering, the ADG was assumed to be affected linearly by heat load. Weekly heat load was calculated as degrees of average temperature-humidity index in excess of a threshold (18 degrees C). The total heat load (H) was the sum of weekly heat loads during the second period. During the months of January to May H was 0; H reached a peak in September. The final BW during the peak of heat stress decreased about 6 kg compared with BW during months of non-heat stress. Weekly and monthly averages of carcass weight generally moved similarly to H. However, there were large fluctuations unrelated to H; the fluctuations were different on the 2 farms. The model included the effects of farm-year of slaughter, sex, age at slaughter, and H, where age at slaughter and H were linear regressions. In analyses, the threshold was varied from 16 to 20 degrees C, and the second period was varied from 8 to 16 wk. The greatest R(2) (10.4%) was at the threshold of temperature-humidity index = 18 degrees C for a period of 10 wk. Varying the threshold and the length of time reduced R(2) less than 1%. Least squares means of year-month and year-week of carcass weight were calculated using a model with the fixed effects farm-year-month or farm-year-week of slaughter, sex, and age at slaughter (linear covariate), and the random effect of birth litter. Changes in BW of finisher pigs due to heat stress can be quantified by H during the last 10 wk of the life of the pig.

The objective of this study was to describe genetic variability of pig carcass weight as a function of heat stress. Data included carcass weights of 23,556 crossbred pigs [Duroc x (Landrace x Large White)] raised on 2 farms in North Carolina and harvested from May 2005 through December 2006. Weather data were obtained from a weather station located about 20 km from the furthest farm. Weekly heat load was calculated as degrees of average temperature-humidity index (THI) in excess of 18 degrees C. The total heat load (H) was the sum of heat loads for 10 wk before harvest. Variance components were estimated with 3 models: univariate (UNI)-not accounting for heat stress, 2-trait (MT2), and random regression (RR). In all of the models, effects included contemporary group, sex, age at harvest, sire, and litter. In MT2, observations in months in which heat stress was observed ("hot") and not observed ("cold") were treated as separate traits. Heat stress was observed in the months of August to November 2005, as well as July to October 2006. No heat stress was observed in the months of May to July 2005, January to June 2006, and November to December 2006. The RR model added a random regression on heat load for the sire effect. Heat load was adjusted to a scale ranging from 0 (no heat stress) to 5 (greatest heat stress). The heritability estimate +/- SE of carcass weight in UNI was 0.17 +/- 0.01. In MT2, the estimates were 0.14 +/- 0.01 for "cold" and 0.28 +/- 0.01 for "hot"; the genetic correlation between carcass weight in "hot" and "cold" months was 0.42 +/- 0.13. The heritability estimates obtained with RR were 0.20 +/- 0.11, 0.19 +/- 0.15, and 0.51 +/- 0.17 for H = 0, 2.5, and 5, respectively. The genetic correlation between the performance in "cold" months (H = 0), and performance under maximum heat load (H = 5) was 0.02, between H = 0 and intermediate heat load (H = 2.5) was 0.52, and between H = 2.5 and H = 5 was 0.86. Rank correlations between EPD derived from the different models ranged from 0.82 to 0.94 between carcass weights under similar H, 0.18 to 0.54 between carcass weights under high and low H, and 0.66 to 0.91 between carcass weights of intermediate and high/low H. Heritability for growth was greater under heat stress. Selection for crossbred performance would be optimal when data for periods both in the absence and presence of heat stress were considered.

The aim of this study was to estimate the genetic correlations between 2 purebred Duroc pig populations (P1 and P2) and their terminal crossbreds [C1 = P1 x (Landrace x Large White) and C2 = P2 x (Landrace x Large White)] raised in different production environments. The traits analyzed were backfat (BF), muscle depth (MD), BW at slaughter (WGT), and weight per day of age (WDA). Data sets from P1, P2, C1, and C2 included 26,674, 8,266, 16,806, and 12,350 animals, respectively. Two-trait models (nucleus and commercial crossbreds) for each group included fixed (contemporary group, sex, weight, and age), random additive (animal for P1 and P2 and sire for C1 and C2), random litter, and random dam (C1 and C2 only) effects. Heritability estimates (+/-SE) for BF were 0.46 +/- 0.04, 0.38 +/- 0.02, 0.32 +/- 0.02, and 0.33 +/- 0.02 for P1, P2, C1, and C2, respectively. Heritability estimates for MD were 0.31 +/- 0.01, 0.23 +/- 0.02, 0.19 +/- 0.01, and 0.12 +/- 0.01 for P1, P2, C1, and C2, respectively. The estimates for WGT and WDA were 0.31 +/- 0.01, 0.21 +/- 0.02, 0.16 +/- 0.01, and 0.18 +/- 0.01 and 0.32 +/- 0.01, 0.22 +/- 0.02, 0.16 +/- 0.01, and 0.19 +/- 0.01, respectively. Genetic correlations between purebreds and crossbreds for BF were 0.83 +/- 0.09 (P1 x C1) and 0.89 +/- 0.05 (P2 x C2), for MD 0.78 +/- 0.05 (P1 x C1) and 0.80 +/- 0.08 (P2 x C2). For WGT and WDA, the correlations were 0.53 +/- 0.08 (P1 x C1), 0.80 +/- 0.10 (P2 x C2), and 0.60 +/- 0.07 (P1 x C1) and 0.79 +/- 0.09 (P2 x C2), respectively. (Co)variances in crossbreds were adjusted to a live BW scale. Compared with purebreds, the genetic variances in crossbreds were lower, and the residual variances were greater. Sire variances in crossbreds were approximately 20 to 30% of the animal variances in purebreds for BF and MD but were 13 to 25% for WGT and WDA. The efficiency of purebred selection on crossbreds, assessed by EBV prediction weights, ranged from 0.43 to 0.91 for line 1 and 0.70 to 0.92 for line 2. When nucleus and commercial environments differ substantially, the efficiency of selection varies by line and traits, and selection strategies that include crossbred data from typical production environments may therefore be desirable.

To study the genetic relationship between three grouped reasons for sow removal (SR) in consecutive parities, accounting for censoring, 13,838 records from Large White sows were analyzed. Data were from seven pure-line farms having, on average, 5.9% unknown SR. Three traits were subjectively defined, each corresponding to a classification of SR (reproductive [RR], nonreproductive [RN], and others [RO]). Records for each trait could take one of five categories, according to parity at removal (0 to 4 or later). A multivariate linear censored model was implemented. The model to estimate (co)variance components and parameters included the effects of year-season, region, contemporary group, and additive genetic effects. The most common SR was related to reproduction (48.5%). Diseases of different origin and cause, old age/parity, and sow death or loss accounted for about 18, 7, and 4% of total culls, respectively. Estimates of variance components showed heterogeneity of additive genetic and residual variances for the three traits. Estimates of heritability were 0.18, 0.13, and 0.15 for RR, RN, and RO, respectively. Genetic correlations between removal codes were high (> or =0.90). Results suggest sizeable additive genetic variances exist for parity at removal and different codes of removal. Different SR reasons seem to operate similarly or as a closely related genetic trait associated with fitness. In particular, RN and RO seem to be genetically indistinguishable. Data structure, definition, and volume are major limitations in studies of sow survival. A multiple-trait censored model is preferred to evaluate reasons of sow disposal. Grouped removal causes seem to be strongly genetically correlated but with heterogeneous variances, suggesting that combining all removal causes and treating the trait as parity at disposal is an alternative approach.

Records of on-test ADG of Large White gilts were analyzed to estimate variance components of direct and associative genetic effects. Models included the effects of contemporary group (farm-barn-batch), birth litter, pen group, and direct and associative additive genetic effects. The area of each pen was 14 m2. The additive genetic variance was a function of the number of competitors in a group, the additive relationships between the animal performing the record and its pen mates, and the additive relationships between pen mates. To partially account for differences in the number of pen mates, a covariable (qi = 1, 1/n, or 1/n(1/2)) was added to the associative genetic effect. There were 4,946 records from 2,409 litters and 362 pen groups. Pen group size ranged from 12 to 16 gilts. Analyses by REML converged very slowly. A grid search showed that the likelihood function was almost flat when the additive genetic associative effect was fitted. Estimates of direct and associative heritability were 0.15 and 0.03, respectively. Within the BLUPF90 family of programs, the mixed-model equations can be set up directly. For variance component estimation, simple programs (REMLF90 and GIBBSF90) worked without modifications, but more optimized programs did not. Estimates obtained using the three values of qi were similar. With the data structure available for this study and under an environment with relative low competition among animals, accurate estimation of associative genetic effects was not possible. Estimation of competitive effects with large pen size is difficult. The magnitude of competition effects may be larger in commercial populations, where housing is denser and food is limited.

Up to 109,447 records of 49,656 Large White sows were used to evaluate the genetic relationship between number of pigs born dead (BD) and number born alive (BA) in first and later parities. Performance data (n = 30,832) for ultrasound backfat (BF) at the end of the test and days to reach 113.5 kg (AD) were used to estimate their relationships with BD and BA at first parity in a four-trait threshold-linear analysis (TL). Effects were year-farm, contemporary group (CG: farm-farrowing year-farrowing month) and animal additive genetic. At first parity, estimates of heritability were 0.09, 0.09, 0.37, and 0.31 for BA, BD, AD, and BF, respectively. The estimate of genetic correlation between BD and litter size was -0.04 (BD-BA). Corresponding values with test traits were both -0.14 (BD-AD, BD-BF). Estimates of genetic correlation between BA and performance traits were 0.08 (BA-AD) and 0.05 (BA-BF). The two test traits were moderately negatively correlated (-0.22). For later parities, a six-trait (BD, BA in three parities) TL model was implemented. The estimates of additive genetic variances and heritability increased with parity for BD and BA. Estimates of heritabilities were: 0.09, 0.10, and 0.11 for BD, and 0.09, 0.12, and 0.12 for BA in parities one to three, respectively. Estimates of genetic correlations between different parities were high (0.91 to 0.96) for BD, and slightly lower (0.74 to 0.95) for BA. Genetic correlations between BD and BA were low and positive (0.02 to 0.17) for BA in Parities 1 and 2, but negative (-0.04 to -0.10) for BA in Parity 3. Selection for increased litter size should have little effect on farrowing piglet mortality. Intense selection for faster growth and increased leanness should increase farrowing piglet mortality of first-parity sows. A repeatability model with a simple correction for the heterogeneity of variances over parities could be implemented to select against farrowing mortality. The genetic components of perinatal piglet mortality are independent of the ones for litter size in the first parity, and they show an undesirable, but not strong, genetic association in second parity.