Bayesian regression models outperform partial least squares methods for predicting milk components and technological properties using infrared spectral data

Abstract

The aim of this study was to assess the performance of Bayesian models commonly used for genomic selection to predict “difficult-to-predict” dairy traits, such as milk fatty acid (FA) expressed as percentage of total fatty acids, and technological properties, such as fresh cheese yield and protein recovery, using Fourier-transform infrared (FTIR) spectral data. Our main hypothesis was that Bayesian models that can estimate shrinkage and perform variable selection may improve our ability to predict FA traits and technological traits above and beyond what can be achieved using the current calibration models (e.g., partial least squares, PLS). To this end, we assessed a series of Bayesian methods and compared their prediction performance with that of PLS. The comparison between models was done using the same sets of data (i.e., same samples, same variability, same spectral treatment) for each trait. Data consisted of 1,264 individual milk samples collected from Brown Swiss cows for which gas chromatographic FA composition, milk coagulation properties, and cheese-yield traits were available. For each sample, 2 spectra in the infrared region from 5,011 to 925 cm−1 were available and averaged before data analysis. Three Bayesian models: Bayesian ridge regression (Bayes RR), Bayes A, and Bayes B, and 2 reference models: PLS and modified PLS (MPLS) procedures, were used to calibrate equations for each of the traits. The Bayesian models used were implemented in the R package BGLR (http://cran.r-project.org/web/packages/BGLR/index.html), whereas the PLS and MPLS were those implemented in the WinISI II software (Infrasoft International LLC, State College, PA). Prediction accuracy was estimated for each trait and model using 25 replicates of a training-testing validation procedure. Compared with PLS, which is currently the most widely used calibration method, MPLS and the 3 Bayesian methods showed significantly greater prediction accuracy. Accuracy increased in moving from calibration to external validation methods, and in moving from PLS and MPLS to Bayesian methods, particularly Bayes A and Bayes B. The maximum R2 value of validation was obtained with Bayes B and Bayes A. For the FA, C10:0 (% of each FA on total FA basis) had the highest R2 (0.75, achieved with Bayes A and Bayes B), and among the technological traits, fresh cheese yield R2 of 0.82 (achieved with Bayes B). These 2 methods have proven to be useful instruments in shrinking and selecting very informative wavelengths and inferring the structure and functions of the analyzed traits. We conclude that Bayesian models are powerful tools for deriving calibration equations, and, importantly, these equations can be easily developed using existing open-source software. As part of our study, we provide scripts based on the open source R software BGLR, which can be used to train customized prediction equations for other traits or populations.

Genetic and environmental relationships of detailed milk fatty acids profile determined by gas chromatography in Brown Swiss cows

“…The heritability estimates reported in these GC-based studies (Bobe et al, 2008;Garnsworthy et al, 2010;Bilal et al, 2014) are comparable to those found in the current paper or somewhat higher (Stoop et al, 2008;Schennink et al, 2008). In contrast, the results from studies based on FTIR predictions (Soyeurt et al, 2007;Bastin et al, 2011Bastin et al, , 2013 have commonly been expressed in terms of the weight of a single fatty acid or group of fatty acids per unit of milk (kg or L) due to the low accuracy of FTIR-based predictions of milk fat content (Ferragina et al, 2015). Data expressed as weight of single or groups of fatty acids per unit of milk would produce heritability estimates for fatty acids similar to those of the milk fat content and often greater than those found for proportion of individual fatty acids on total fatty acid content, as in the current and other GC-based experiments.…”

A landscape of the heritability of Fourier-transform infrared spectral wavelengths of milk samples by parity and lactation stage in Holstein cows

“…In the sector of the wavenumbers 2,862 to 3,000 cm −1 , several contiguous heritability peaks appear that are well defined in third and later months of lactation. This region has been associated with fat yield (e.g., Bittante and Cecchinato, 2013) and Fat% (e.g., Wang and Bovenhuis, 2018) and with specific fatty acids [e.g., palmitic (C16:0); Ferragina et al, 2015]. Interestingly, the same wavenumbers that were characterized by high heritability estimates in mo 3 to 4 also had very low estimated heritabilities in the first month of the lactation (Figure 3).…”

Genetic and environmental relationships of different measures of individual cheese yield and curd nutrients recovery with coagulation properties of bovine milk

“…Although breeding values for CY and REC have been obtained from direct measures of phenotypes in experimental studies (Bittante et al, 2013a), selection strategies based on direct measures of CY and REC would be very hard to implement. High throughput (potentially, hundreds of samples per hour), ease of use, and reduced cost of analysis make Fourier transform infrared (FTIR) spectroscopy a promising alternative for the assessment of CY and REC at the population level Ferragina et al, 2015).…”