Integrating genomic and infrared spectral data improves the prediction of milk protein composition in dairy cattle

Abstract

Background Over the past decade, Fourier transform infrared (FTIR) spectroscopy has been used to predict novel milk protein phenotypes. Genomic data might help predict these phenotypes when integrated with milk FTIR spectra. The objective of this study was to investigate prediction accuracy for milk protein phenotypes when heterogeneous on-farm, genomic, and pedigree data were integrated with the spectra. To this end, we used the records of 966 Italian Brown Swiss cows with milk FTIR spectra, on-farm information, medium-density genetic markers, and pedigree data. True and total whey protein, and five casein, and two whey protein traits were analyzed. Multiple kernel learning constructed from spectral and genomic (pedigree) relationship matrices and multilayer BayesB assigning separate priors for FTIR and markers were benchmarked against a baseline partial least squares (PLS) regression. Seven combinations of covariates were considered, and their predictive abilities were evaluated by repeated random sub-sampling and herd cross-validations (CV). Results Addition of the on-farm effects such as herd, days in milk, and parity to spectral data improved predictions as compared to those obtained using the spectra alone. Integrating genomics and/or the top three markers with a large effect further enhanced the predictions. Pedigree data also improved prediction, but to a lesser extent than genomic data. Multiple kernel learning and multilayer BayesB increased predictive performance, whereas PLS did not. Overall, multilayer BayesB provided better predictions than multiple kernel learning, and lower prediction performance was observed in herd CV compared to repeated random sub-sampling CV. Conclusions Integration of genomic information with milk FTIR spectral can enhance milk protein trait predictions by 25% and 7% on average for repeated random sub-sampling and herd CV, respectively. Multiple kernel learning and multilayer BayesB outperformed PLS when used to integrate heterogeneous data for phenotypic predictions.

Real-time milk analysis integrated with stacking ensemble learning as a tool for the daily prediction of cheese-making traits in Holstein cattle

“…Prediction performance here is directly affected by the cross-validation strategy and the biological characteristics of the target phenotype (Mota et al, 2021b). Several authors have observed that training and validation sets may overlap to a greater (k-fold) or lesser extent (leave-one-herd-out or independent validation), confirmation that lower dependencies correspond to lower prediction accuracies and greater prediction errors (Meyer et al, 2018;Wang and Bovenhuis, 2019;Baba et al, 2021;Mota et al, 2021a). It is important to note that stacking ensemble learning was the technique exhibiting the smallest reduction in predictive ability when the dependence between the training and validation sets was reduced.…”

“…Infrared predictions for cheese-making efficiency at the cow level are useful to aid farm management decisions, whereas advances in high-throughput phenotyping techniques bridge the genotype-to-phenotype gap in dairy cattle selection. Infrared technology, and particularly Fourier-transform infrared, has been widely used to predict several phenotypes in dairy cattle, including cheese-making traits (Kaniyamattam and De Vries, 2014;Baba et al, 2021;Mota et al, 2021a). Recently, near-infrared spectroscopy (NIR) has also started to be used with in-line on-farm applications, such as the AfiLab real-time milk analyzer (Afimilk), which has been proved to provide a robust, on-farm data stream for milk quality traits.…”

“…On-farm data generated by high-throughput technologies and multiple sources of information from each animal (DIM, parity, health status, and so on) can be integrated with infrared information to predict novel phenotypes, which could be exploited for selective breeding or management purposes (or both). In recent years, increasing attention has been directed toward on-farm data collection and integration for predicting phenotypic traits of economic importance to the dairy cattle industry (Baba et al, 2021). Including DIM in the AfiLab infrared predictions (model 1) increased the R 2 by around 5.8% for traditional MCP, 8.5% for CF t , 5.3% for CY, and 5.9% for REC (Supplemental Figure S5; https: / / doi .org/ 10 .6084/ m9 .figshare .18197720 .v1, Mota et al, 2022).…”

Classification of plant growth-promoting bacteria inoculation status and prediction of growth-related traits in tropical maize using hyperspectral image and genomic data

“…Multiomic prediction was performed by combining hyperspectral and genomic data using BayesRR and BayesB for growth-related phenotypes by setting different priors for each omic covariate. The framework closely followed that of Gonçalves et al (2021) and Baba et al (2021). The mixture parameter π was set to 0.99 for hyperspectral and genomic terms in multi-omics BayesB.…”

“…The framework closely followed that of Gonçalves et al (2021) and Baba et al (2021). The mixture parameter π was set to 0.99 for hyperspectral and genomic terms in multi-omics BayesB.…”