Would large dataset sample size unveil the potential of deep neural networks for improved genome-enabled prediction of complex traits? The case for body weight in broilers.

BACKGROUND: Deep neural networks (DNN) are a particular case of artificial neural networks (ANN) composed by multiple hidden layers, and have recently gained attention in genome-enabled prediction of complex traits. Yet, few studies in genome-enabled prediction have assessed the performance of DNN compared to traditional regression models. Strikingly, no clear superiority of DNN has been reported so far, and results seem highly dependent on the species and traits of application. Nevertheless, the relatively small datasets used in previous studies, most with fewer than 5000 observations may have precluded the full potential of DNN. Therefore, the objective of this study was to investigate the impact of the dataset sample size on the performance of DNN compared to Bayesian regression models for genome-enable prediction of body weight in broilers by sub-sampling 63,526 observations of the training set. RESULTS: Predictive performance of DNN improved as sample size increased, reaching a plateau at about 0.32 of prediction correlation when 60% of the entire training set size was used (i.e., 39,510 observations). Interestingly, DNN showed superior prediction correlation using up to 3% of training set, but poorer prediction correlation after that compared to Bayesian Ridge Regression (BRR) and Bayes Cπ. Regardless of the amount of data used to train the predictive machines, DNN displayed the lowest mean square error of prediction compared to all other approaches. The predictive bias was lower for DNN compared to Bayesian models, across all dataset sizes, with estimates close to one with larger sample sizes. CONCLUSIONS: DNN had worse prediction correlation compared to BRR and Bayes Cπ, but improved mean square error of prediction and bias relative to both Bayesian models for genome-enabled prediction of body weight in broilers. Such findings, highlights advantages and disadvantages between predictive approaches depending on the criterion used for comparison. Furthermore, the inclusion of more data per se is not a guarantee for the DNN to outperform the Bayesian regression methods commonly used for genome-enabled prediction. Nonetheless, further analysis is necessary to detect scenarios where DNN can clearly outperform Bayesian benchmark models.

Improving genomic prediction accuracy for meat tenderness in Nellore cattle using artificial neural networks.

The goal of this study was to compare the predictive performance of artificial neural networks (ANNs) with Bayesian ridge regression, Bayesian Lasso, Bayes A, Bayes B and Bayes Cπ in estimating genomic breeding values for meat tenderness in Nellore cattle. The animals were genotyped with the Illumina Bovine HD Bead Chip (HD, 777K from 90 samples) and the GeneSeek Genomic Profiler (GGP Indicus HD, 77K from 485 samples). The quality control for the genotypes was applied on each Chip and comprised removal of SNPs located on non-autosomal chromosomes, with minor allele frequency <5%, deviation from HWE (p < 10-6 ), and with linkage disequilibrium >0.8. The FImpute program was used for genotype imputation. Pedigree-based analyses indicated that meat tenderness is moderately heritable (0.35), indicating that it can be improved by direct selection. Prediction accuracies were very similar across the Bayesian regression models, ranging from 0.20 (Bayes A) to 0.22 (Bayes B) and 0.14 (Bayes Cπ) to 0.19 (Bayes A) for the additive and dominance effects, respectively. ANN achieved the highest accuracy (0.33) of genomic prediction of genetic merit. Even though deep neural networks are recognized to deliver more accurate predictions, in our study ANN with one single hidden layer, 105 neurons and rectified linear unit (ReLU) activation function was sufficient to increase the prediction of genetic merit for meat tenderness. These results indicate that an ANN with relatively simple architecture can provide superior genomic predictions for meat tenderness in Nellore cattle.

The effect of life history events on carcass merit and price of cull dairy cows.

Cull dairy cows contribute almost 10% of national beef production in the United States. However, different factors throughout the life of dairy cows affect their weight and overall body condition as well as carcass traits, and consequently affect their market price. Therefore, the objectives of this study were: (1) to assess relationships between price ratio and carcass merit of cull dairy cows sold through several sites of an auction market and (2) to investigate the effect of animal life history events and live weight on sale barn price (BP) and price ratio (as a measure of relative price), as an indicator of carcass merit. Data from 4 dairy operations included 3,602 cull dairy cow records during the period of 2015 to 2019. Life history events data were collected from each dairy operation through Dairy Comp software; live weight and price were obtained periodically from the auction market, and the carcass data were provided by a local packing plant. Cow price in dollars per unit of live weight ($/cwt) and price ratio were the 2 outcome variables used in the analyses. Price ratio was created aiming to remove seasonality effects from BP (BP divided by the national average price for its respective month and year of sale). The association between price ratio and carcass merit traits was investigated using canonical correlation analysis, and the effect of life history events on both BP and price ratio was inferred using a multiple linear regression technique. More than 70% of the cows were culled in the first 3 lactations, with an average live weight of 701.5 kg, carcass weight of 325 kg, and dressing percentage of 46.3%. On average, cull cows were sold at $57.0/cwt during the period considered. The canonical correlation between price ratio and carcass merit traits was 0.76, indicating that price ratio reflected carcass merit of cull cows. Later lactations led to lower BP compared with cows culled during the first 2 lactations. Injury, and leg and feet problems negatively affected BP. Productive variables demonstrated that the greater milk production might lead to lower cow prices. A large variation between farms was also noted. In conclusion, price ratio was a good indicator of carcass merit of cull cows, and life history events significantly affected sale BP and carcass merit of cull cows sold through auction markets.

Network analysis of swine movements in a multi-site pig production system in Iowa, USA.

Pig production in the United States is based on multi-site systems in which pigs are transported between farms after the conclusion of each particular production phase. Although ground transportation is a critical component of the pork supply chain, it might constitute a potential route of infectious disease dissemination. Here, we used a time series network analysis to: (1) describe pig movement flow in a multi-site production system in Iowa, USA, (2) conduct percolation analysis to investigate network robustness to interventions for diseases with different transmissibility, and (3) assess the potential impact of each farm type on disease dissemination across the system. Movement reports from 2014-2016 were provided by Iowa Select Farms, Iowa Fall, IA. A total of 76,566 shipments across sites was analyzed, and time series network analyses with temporal resolution of 1, 3, 6, 12, and 36 months were considered. The general topological properties of networks with resolution of 1, 3, 6, and 12 months were compared with the whole period static network (36 months) and included the following features: number of nodes and edges, degree assortativity, density, average path length, diameter, clustering coefficients, giant strongly connected component, giant weakly connected component, giant in component, and giant out component. Small-world and scale-free topologies, centrality parameters, and percolation analysis were investigated for the networks with 1-month window. Networks' robustness to interventions was assessed by using the Basic Reproduction Number (R0). Centrality parameters indicate that gilt development units (GDU), nursery, and sow farms have more central role in the pig production hierarchical structure. Therefore, they are potentially major factors of introduction and spread of diseases over the system. Wean-to-finishing and finishing sites displayed high in-degree values, indicating that they are more susceptible to be infected. Percolation analysis combined with general properties (i.e. heavy-tailed distributions and degree disassortative) suggested that networks with 1-month time resolution were highly responsive to interventions. Furthermore, the characteristics of a disease should have strong implications in the biosecurity practices across production sites. For instance, biosecurity practices should be focused on sow farms for highly contagious disease (e.g., foot and mouth disease), while it should target nursery sites in the case of a less contagious diseases (i.e. mycobacterial infections). Understanding the patterns of swine movements is crucial for the swine industry decision-making in the case of an epidemic, as well as to design cost-effective approaches to monitor, prevent, control and eradicate infectious diseases in multi-site systems.

Generalized additive mixed model on the analysis of total transport losses of market-weight pigs1.

Transportation losses of market-weight pigs are an animal welfare concern, and result in direct economic impact for producers and abattoirs. Such losses are related to multiple factors including pig genetics, human handling, management, and weather conditions. Understanding the factors associated with total transport losses (TTL) is important to the swine industry because it can aid decision-making, and help in the development of transportation strategies to minimize the risk of losses. Hence, the objective of this study was to investigate factors associated with TTL on market-weight pigs in typically field conditions for Midwestern United States using a generalized additive mixed model (GAMM). The final quasi-binomial GAMM included the fixed (main and interactions) effects of abattoir of destination, type of driver, average market weight, distance traveled, wind speed, precipitation, and temperature-humidity index (THI), as well as the random effects of truck companies and the combination of site of origin and period of the year. Results indicate significant associations between TTL and the main effect of all explanatory variables (P < 0.05), except for wind speed and precipitation. Interactions of average market weight × abattoir, and wind speed × precipitation were also significant. A complex nonlinear relationship between TTL and model covariates were observed for distance traveled, THI, and interaction terms. This study showed that TTL of market-weight pigs are caused by a complex system involving multiple interacting factors, which can be potentially managed to mitigate the risk of losses. In addition, the GAMM showed to be a simple and flexible approach to model TTL because it can capture nonlinear relationships, handle non-normal data, and can potentially accommodate data structure.