Preweaning piglet survival on commercial farms.

Preweaning piglet mortality (PWM), a trait highly related to litter size, is one of the main concerns associated with productive efficiency and animal welfare in commercial pig farms. The objectives of this work were to study piglet survival at the farm level, to establish a survival rate (SR) as a target indicator to be improved, and to model it based on other reproductive parameters. Analyzed data corresponded to 580 Spanish commercial farms with a total inventory of 809,768 sows. These farms showed a mean SR of 85.70% piglets born alive (BA), which decreased to 81.81% when total piglets born (TB) were considered. The SR was strongly associated with prolificacy (P < 0.01), the parities with the highest prolificacy being those that had the lowest SR. Thus, the highest correlations were for the SR of piglets BA in the third and fourth parities (r = -0.460 and r = -0.452, respectively, P < 0.01), and for the SR of piglets TB in the fourth parity (r = -0.546, P < 0.01), which was the one with the highest prolificacy. The values corresponding to the quartile of farms with the highest SR within the most productive farms were established as targets to be improved, which were ≥88.5% of piglets BA and 83.2% of piglets TB. Nevertheless, the direct associations shown between the piglet's survival and prolificacy and other productive factors, such as the age of piglets at weaning, the farrowings per sow and year and the farrowing interval, suggest the convenience of modeling the risk of PWM on farms to have its own target of survival index to be improved.

Sow prolificacy and preweaning piglet mortality have increased parallelly on commercial farms. This loss of piglets is a concern of efficiency and animal welfare, and it requires the improvement of piglet survival by reducing the number of stillborn piglets and piglet mortality during lactation, paying particular attention to hyperprolific sows (≥15 total piglets born per litter). Data from 580 commercial farms with an average inventory of 809,768 sows have been analyzed to propose two predictive models based on several reproductive parameters and two survival rate targets with the aim of reducing this problem, which are ≥88.5% of piglets born alive and ≥83.2% of total piglets born.

Impact of multi-output and stacking methods on feed efficiency prediction from genotype using machine learning algorithms.

Feeding represents the largest economic cost in meat production; therefore, selection to improve traits related to feed efficiency is a goal in most livestock breeding programs. Residual feed intake (RFI), that is, the difference between the actual and the expected feed intake based on animal's requirements, has been used as the selection criteria to improve feed efficiency since it was proposed by Kotch in 1963. In growing pigs, it is computed as the residual of the multiple regression model of daily feed intake (DFI), on average daily gain (ADG), backfat thickness (BFT), and metabolic body weight (MW). Recently, prediction using single-output machine learning algorithms and information from SNPs as predictor variables have been proposed for genomic selection in growing pigs, but like in other species, the prediction quality achieved for RFI has been generally poor. However, it has been suggested that it could be improved through multi-output or stacking methods. For this purpose, four strategies were implemented to predict RFI. Two of them correspond to the computation of RFI in an indirect way using the predicted values of its components obtained from (i) individual (multiple single-output strategy) or (ii) simultaneous predictions (multi-output strategy). The other two correspond to the direct prediction of RFI using (iii) the individual predictions of its components as predictor variables jointly with the genotype (stacking strategy), or (iv) using only the genotypes as predictors of RFI (single-output strategy). The single-output strategy was considered the benchmark. This research aimed to test the former three hypotheses using data recorded from 5828 growing pigs and 45,610 SNPs. For all the strategies two different learning methods were fitted: random forest (RF) and support vector regression (SVR). A nested cross-validation (CV) with an outer 10-folds CV and an inner threefold CV for hyperparameter tuning was implemented to test all strategies. This scheme was repeated using as predictor variables different subsets with an increasing number (from 200 to 3000) of the most informative SNPs identified with RF. Results showed that the highest prediction performance was achieved with 1000 SNPs, although the stability of feature selection was poor (0.13 points out of 1). For all SNP subsets, the benchmark showed the best prediction performance. Using the RF as a learner and the 1000 most informative SNPs as predictors, the mean (SD) of the 10 values obtained in the test sets were: 0.23 (0.04) for the Spearman correlation, 0.83 (0.04) for the zero-one loss, and 0.33 (0.03) for the rank distance loss. We conclude that the information on predicted components of RFI (DFI, ADG, MW, and BFT) does not contribute to improve the quality of the prediction of this trait in relation to the one obtained with the single-output strategy.

Genic and non-genic SNP contributions to additive and dominance genetic effects in purebred and crossbred pig traits.

The present research has estimated the additive and dominance genetic variances of genic and intergenic segments for average daily gain (ADG), backfat thickness (BFT) and pH of the semimembranosus dorsi muscle (PHS). Further, the predictive performance using additive and additive dominance models in a purebred Piétrain (PB) and a crossbred (Piétrain × Large White, CB) pig population was assessed. All genomic regions contributed equally to the additive and dominance genetic variations and lead to the same predictive ability that did not improve with the inclusion of dominance genetic effect and inbreeding in the models. Using all SNPs available, additive genotypic correlations between PB and CB performances for the three traits were high and positive (> 0.83) and dominance genotypic correlation was very inaccurate. Estimates of dominance genotypic correlations between all pairs of traits in both populations were imprecise but positive for ADG-BFT in CB and BFT-PHS in PB and CB with a high probability (> 0.98). Additive and dominance genotypic correlations between BFT and PHS were of different sign in both populations, which could indicate that genes contributing to the additive genetic progress in both traits would have an antagonistic effect when used for exploiting dominance effects in planned matings.

Feature Selection Stability and Accuracy of Prediction Models for Genomic Prediction of Residual Feed Intake in Pigs Using Machine Learning.

Feature selection (FS, i.e., selection of a subset of predictor variables) is essential in high-dimensional datasets to prevent overfitting of prediction/classification models and reduce computation time and resources. In genomics, FS allows identifying relevant markers and designing low-density SNP chips to evaluate selection candidates. In this research, several univariate and multivariate FS algorithms combined with various parametric and non-parametric learners were applied to the prediction of feed efficiency in growing pigs from high-dimensional genomic data. The objective was to find the best combination of feature selector, SNP subset size, and learner leading to accurate and stable (i.e., less sensitive to changes in the training data) prediction models. Genomic best linear unbiased prediction (GBLUP) without SNP pre-selection was the benchmark. Three types of FS methods were implemented: (i) filter methods: univariate (univ.dtree, spearcor) or multivariate (cforest, mrmr), with random selection as benchmark; (ii) embedded methods: elastic net and least absolute shrinkage and selection operator (LASSO) regression; (iii) combination of filter and embedded methods. Ridge regression, support vector machine (SVM), and gradient boosting (GB) were applied after pre-selection performed with the filter methods. Data represented 5,708 individual records of residual feed intake to be predicted from the animal's own genotype. Accuracy (stability of results) was measured as the median (interquartile range) of the Spearman correlation between observed and predicted data in a 10-fold cross-validation. The best prediction in terms of accuracy and stability was obtained with SVM and GB using 500 or more SNPs [0.28 (0.02) and 0.27 (0.04) for SVM and GB with 1,000 SNPs, respectively]. With larger subset sizes (1,000-1,500 SNPs), the filter method had no influence on prediction quality, which was similar to that attained with a random selection. With 50-250 SNPs, the FS method had a huge impact on prediction quality: it was very poor for tree-based methods combined with any learner, but good and similar to what was obtained with larger SNP subsets when spearcor or mrmr were implemented with or without embedded methods. Those filters also led to very stable results, suggesting their potential use for designing low-density SNP chips for genome-based evaluation of feed efficiency.

Purebred and Crossbred Genomic Evaluation and Mate Allocation Strategies To Exploit Dominance in Pig Crossbreeding Schemes.

We investigated the effectiveness of mate allocation strategies accounting for non-additive genetic effects to improve crossbred performance in a two-way crossbreeding scheme. We did this by computer simulation of 10 generations of evaluation and selection. QTL effects were simulated as correlated across purebreds and crossbreds, and (positive) heterosis was simulated as directional dominance. The purebred-crossbred correlation was 0.30 or 0.68 depending on the genetic variance component used. Dominance and additive marker effects were estimated simultaneously for purebreds and crossbreds by multiple trait genomic BLUP. Four scenarios that differ in the sources of information (only purebred data, or purebred and crossbred data) and mate allocation strategies (mating at random, minimizing expected future inbreeding, or maximizing the expected total genetic value of crossbred animals) were evaluated under different cases of genetic variance components. Selecting purebred animals for purebred performance yielded a response of 0.2 genetic standard deviations of the trait "crossbred performance" per generation, whereas selecting purebred animals for crossbred performance doubled the genetic response. Mate allocation strategy to maximize the expected total genetic value of crossbred descendants resulted in a slight increase (0.8%, 4% and 0.5% depending on the genetic variance components) of the crossbred performance. Purebred populations increased homozygosity, but the heterozygosity of the crossbreds remained constant. When purebred-crossbred genetic correlation is low, selecting purebred animals for crossbred performance using crossbred information is a more efficient strategy to exploit heterosis and increase performance at the crossbred commercial level, whereas mate allocation did not improve crossbred performance.

Performance of pedigree and various forms of marker-derived relationship coefficients in genomic prediction and their correlations.

In recent years, with development and validation of different genotyping panels, several methods have been proposed to build efficient similarity matrices among individuals to be used for genomic selection. Consequently, the estimated genetic parameters from such information may deviate from their counterpart using traditional family information. In this study, we used a pedigree-based numerator relationship matrix (A) and three types of marker-based relationship matrices ( G ) including two identical by descent, that is G K and G M and one identical by state, G V as well as four Gaussian kernel ( GK ) similarity kernels with different smoothing parameters to predict yet to be observed phenotypes. Also, we used different kinship matrices that are a linear combination of marker-derived IBD or IBS matrices with A, constructed as K = λ G + 1 - λ A , where the weight ( λ ) assigned to each source of information varied over a grid of values. A Bayesian multiple-trait Gaussian model was fitted to estimate the genetic parameters and compare the prediction accuracy in terms of predictive correlation, mean square error and unbiasedness. Results show that the estimated genetic parameters (heritability and correlations) are affected by the source of the information used to create kinship or the weight placed on the sources of genomic and pedigree information. The superiority of GK-based model depends on the smoothing parameters (θ) so that with an optimum θ value, the GK-based model statistically yielded better performance (higher predictive correlation, lowest MSE and unbiased estimates) and more stable correlations and heritability than the model with IBD, IBS or A kinship matrices or any of the linear combinations.

Machine Learning Prediction of Crossbred Pig Feed Efficiency and Growth Rate From Single Nucleotide Polymorphisms.

This research assessed the ability of a Support Vector Machine (SVM) regression model to predict pig crossbred (CB) performance from various sources of phenotypic and genotypic information for improving crossbreeding performance at reduced genotyping cost. Data consisted of average daily gain (ADG) and residual feed intake (RFI) records and genotypes of 5,708 purebred (PB) boars and 5,007 CB pigs. Prediction models were fitted using individual PB genotypes and phenotypes (trn.1); genotypes of PB sires and average of CB records per PB sire (trn.2); and individual CB genotypes and phenotypes (trn.3). The average of CB offspring records was the trait to be predicted from PB sire's genotype using cross-validation. Single nucleotide polymorphisms (SNPs) were ranked based on the Spearman Rank correlation with the trait. Subsets with an increasing number (from 50 to 2,000) of the most informative SNPs were used as predictor variables in SVM. Prediction performance was the median of the Spearman correlation (SC, interquartile range in brackets) between observed and predicted phenotypes in the testing set. The best predictive performances were obtained when sire phenotypic information was included in trn.1 (0.22 [0.03] for RFI with SVM and 250 SNPs, and 0.12 [0.05] for ADG with SVM and 500-1,000 SNPs) or when trn.3 was used (0.29 [0.16] with Genomic best linear unbiased prediction (GBLUP) for RFI, and 0.15 [0.09] for ADG with just 50 SNPs). Animals from the last two generations were assigned to the testing set and remaining animals to the training set. Individual's PB own phenotype and genotype improved the prediction ability of CB offspring of young animals for ADG but not for RFI. The highest SC was 0.34 [0.21] and 0.36 [0.22] for RFI and ADG, respectively, with SVM and 50 SNPs. Predictive performance using CB data for training leads to a SC of 0.34 [0.19] with GBLUP and 0.28 [0.18] with SVM and 250 SNPs for RFI and 0.34 [0.15] with SVM and 500 SNPs for ADG. Results suggest that PB candidates could be evaluated for CB performance with SVM and low-density SNP chip panels after collecting their own RFI or ADG performances or even earlier, after being genotyped using a reference population of CB animals.

SNP-based mate allocation strategies to maximize total genetic value in pigs.

BACKGROUND: Mate allocation strategies that account for non-additive genetic effects can be used to maximize the overall genetic merit of future offspring. Accounting for dominance effects in genetic evaluations is easier in a genomic context, than in a classical pedigree-based context because the combinations of alleles at loci are known. The objective of our study was two-fold. First, dominance variance components were estimated for age at 100 kg (AGE), backfat depth (BD) at 140 days, and for average piglet weight at birth within litter (APWL). Second, the efficiency of mate allocation strategies that account for dominance and inbreeding depression to maximize the overall genetic merit of future offspring was explored. RESULTS: Genetic variance components were estimated using genomic models that included inbreeding depression with and without non-additive genetic effects (dominance). Models that included dominance effects did not fit the data better than the genomic additive model. Estimates of dominance variances, expressed as a percentage of additive genetic variance, were 20, 11, and 12% for AGE, BD, and APWL, respectively. Estimates of additive and dominance single nucleotide polymorphism effects were retrieved from the genetic variance component estimates and used to predict the outcome of matings in terms of total genetic and breeding values. Maximizing total genetic values instead of breeding values in matings gave the progeny an average advantage of - 0.79 days, - 0.04 mm, and 11.3 g for AGE, BD and APWL, respectively, but slightly reduced the expected additive genetic gain, e.g. by 1.8% for AGE. CONCLUSIONS: Genomic mate allocation accounting for non-additive genetic effects is a feasible and potential strategy to improve the performance of the offspring without dramatically compromising additive genetic gain.

Predictive ability of genome-assisted statistical models under various forms of gene action.

Recent work has suggested that the performance of prediction models for complex traits may depend on the architecture of the target traits. Here we compared several prediction models with respect to their ability of predicting phenotypes under various statistical architectures of gene action: (1) purely additive, (2) additive and dominance, (3) additive, dominance, and two-locus epistasis, and (4) purely epistatic settings. Simulation and a real chicken dataset were used. Fourteen prediction models were compared: BayesA, BayesB, BayesC, Bayesian LASSO, Bayesian ridge regression, elastic net, genomic best linear unbiased prediction, a Gaussian process, LASSO, random forests, reproducing kernel Hilbert spaces regression, ridge regression (best linear unbiased prediction), relevance vector machines, and support vector machines. When the trait was under additive gene action, the parametric prediction models outperformed non-parametric ones. Conversely, when the trait was under epistatic gene action, the non-parametric prediction models provided more accurate predictions. Thus, prediction models must be selected according to the most probably underlying architecture of traits. In the chicken dataset examined, most models had similar prediction performance. Our results corroborate the view that there is no universally best prediction models, and that the development of robust prediction models is an important research objective.

Pedigree and genomic evaluation of pigs using a terminal-cross model.

BACKGROUND: In crossbreeding schemes, within-line selection of purebreds is performed mainly to improve the performance of crossbred descendants under field conditions. The genetic correlation between purebred and crossbred performance is an important parameter to be assessed because purebred performance can be a poor predictor of the performance of crossbred offspring. With the availability of high-density markers, the feasibility of using crossbred information to evaluate purebred candidates can be reassessed. This study implements and applies a single-step terminal-cross model (GEN) to real data to estimate the genetic parameters of several production and quality traits in pigs. METHODS: Piétrain sires were mated with Piétrain and Large White dams to produce purebred and crossbred male half-sib piglets; growth rate, feed conversion ratio, lean meat, pH of longissimus dorsi muscle, drip loss and intramuscular fat content were recorded on all half-sibs. Animals were genotyped using the Illumina Porcine SNP60 BeadChip. The genetic correlation between purebred and crossbred performance was estimated separately for each trait. Purebred animals were evaluated using an animal model, whereas the additive genetic effect of a crossbred individual was decomposed into the additive effects of the sire and dam and a Mendelian sampling effect that was confounded with the residual effect. Genotypes of the Piétrain animals were integrated in the genetic evaluation by using a single-step procedure. As benchmarks, we used a model that was identical to GEN but only accounted for pedigree information (PED) and also two univariate single-step models (GEN_UNI) that took either purebred or crossbred performance into account. RESULTS: Genetic correlations between purebred and crossbred performance were high and positive for all traits (>0.69). Accuracies of estimated breeding values of genotyped sires and purebred offspring that were obtained with the GEN model outperformed both those obtained with the PED and the GEN_UNI models. The use of genomic information increased the predictive ability of the GEN model, but it did not substantially outperform the GEN_UNI models. CONCLUSIONS: We present a single-step terminal-cross model that integrates genomic information of purebred and crossbred performance by using available software. It improves the theoretical accuracy of genetic evaluations in breeding programs that are based on crossbreeding.

Daily exposure to summer circadian cycles affects spermatogenesis, but not fertility in an in vivo rabbit model.

Heat stress (HS) in mammals is a determining factor in the deterioration of spermatogenesis and can cause infertility. The aim of this study was to evaluate the effect of continuous summer circadian cycles on semen production, sperm cell features, fertility, prolificacy, and fecal cortisol metabolites from rabbits kept under an in vivo HS model. We split randomly 60 New Zealand White rabbits into two temperature-controlled rooms: The control group was maintained at comfort temperature (18 °C-22 °C) and an HS group, where the environmental temperature was programmed to increase from 22 °C to 31 °C and be maintained for 3 hours to this temperature at the central part of the day. Fecal cortisol metabolites were assessed to evaluate the stress conditions. Seminal parameters were analyzed. Although animals exposed to HS showed higher values of fecal cortisol metabolites (P = 0.0003), no differences were detected in fertility or prolificacy. Semen samples from HS males showed a significant decrease (P < 0.05) with respect to the controls in the percentage of viable spermatozoa (80.71% vs. 74.21%), and a significant (P ≤ 0.01) increase in the percentage of acrosomic abnormalities (22.57% vs. 36.96%) and tailless spermatozoa (7.91% vs. 12.83). Among motility parameters, no differences were found. This study describes a model of HS simulating a continuous summer daily cycle that allows periods of time to recover as it occurs under natural conditions. Although negative effects have been detected in several sperm parameters, fertility and prolificacy were not affected, suggesting a recovery of the reproductive function when normal conditions are reestablished.