Search

Direct measurements of methane (CH4) from individual animals are difficult and expensive. Predictions based on proxies for CH4 are a viable alternative. Most prediction models are based on multiple linear regressions (MLR) and predictor variables that are not routinely available in commercial farms, such as dry matter intake (DMI) and diet composition. The use of machine learning (ML) algorithms to predict CH4 emissions from across-country heterogeneous data sets has not been reported. The objectives were to compare performances of ML ensemble algorithm random forest (RF) and MLR models in predicting CH4 emissions from proxies in dairy cows, and assess effects of imputing missing data points on prediction accuracy. Data on CH4 emissions and proxies for CH4 from 20 herds were provided by 10 countries. The integrated data set contained 43,519 records from 3,483 cows, with 18.7% missing data points imputed using k-nearest neighbor imputation. Three data sets were created, 3k (no missing records), 21k (missing DMI imputed from milk, fat, protein, body weight), and 41k (missing DMI, milk fat, and protein records imputed). These data sets were used to test scenarios (with or without DMI, imputed vs. nonimputed DMI, milk fat, and protein), and prediction models (RF vs. MLR). Model predictive ability was evaluated within and between herds through 10-fold cross-validation. Prediction accuracy was measured as correlation between observed and predicted CH4, root mean squared error (RMSE) and mean normalized discounted cumulative gain (NDCG). Inclusion of DMI in the model improved within and between-herd prediction accuracy to 0.77 (RMSE = 23.3%) and 0.58 (RMSE = 31.9%) in RF and to 0.50 (RMSE = 0.327) and 0.13 (RMSE = 42.71) in MLR, respectively than when DMI was not included in the predictive model. When missing DMI records were imputed, within and between-herd accuracy increased to 0.84 (RMSE = 18.5%) and 0.63 (RMSE = 29.9%), respectively. In all scenarios, RF models out-performed MLR models. Results suggest routinely measured variables from dairy farms can be used in developing globally robust prediction models for CH4 if coupled with state-of-the-art techniques for imputation and advanced ML algorithms for predictive modeling., Journal of Dairy Science, 105 (6), ISSN:0022-0302, ISSN:1525-3198

info:eu-repo/semantics/inPress

Simple Summary Dry matter intake, related to the number of nutrients available to an animal to meet its production and health needs, is crucial for the economic, environmental, and welfare management of dairy herds. Because the equipment required to weigh the ingested food at an individual level is not broadly available, we propose some new ways to approach the actual dry matter consumed by a dairy cow for a given day. To do so, we used regression models using parity (number of lactations), week of lactation, milk yield, milk mid-infrared spectrum, and prediction of bodyweight, fat, protein, lactose, and fatty acids content in milk. We chose these elements to predict individual dry matter intake because they are either easily accessible or routinely provided by regional dairy organizations (often called “dairy herd improvement” associations). We succeeded in producing a model whose dry matter intake predictions were moderately related to the actual values. Abstract We predicted dry matter intake of dairy cows using parity, week of lactation, milk yield, milk mid-infrared (MIR) spectrum, and MIR-based predictions of bodyweight, fat, protein, lactose, and fatty acids content in milk. The dataset comprised 10,711 samples of 534 dairy cows with a geographical diversity (Australia, Canada, Denmark, and Ireland). We set up partial least square (PLS) regressions with different constructs and a one-hidden-layer artificial neural network (ANN) using the highest contribution variables. In the ANN, we replaced the spectra with their projections to the 25 first PLS factors explaining 99% of the spectral variability to reduce the model complexity. Cow-independent 10 × 10-fold cross-validation (CV) achieved the best performance with root mean square errors (RMSECV) of 3.27 ± 0.08 kg for the PLS regression and 3.25 ± 0.13 kg for ANN. Although the available data were significantly different, we also performed a country-independent validation (CIV) to measure the models’ performance fairly. We found RMSECIV varying from 3.73 to 6.03 kg for PLS and 3.69 to 5.08 kg for ANN. Ultimately, based on the country-independent validation, we discussed the developed models’ performance with those achieved by the National Research Council’s equation.

The objective of this study was to estimate genetic parameters of predicted N use efficiency (PNUE) and N losses (PNL) as proxies of N use and loss for Holstein cows. Furthermore, we have assessed approximate genetic correlations between PNUE, PNL, and dairy production, health, longevity, and conformation traits. These traits are considered important in many countries and are currently evaluated by the International Bull Evaluation Service (Interbull). The values of PNUE and PNL were obtained by using the combined milk mid-infrared (MIR) spectrum, parity, and milk yield–based prediction equations on test-day MIR records with days in milk (DIM) between 5 and 50 d. After editing, the final data set comprised 46,163 records of 21,462 cows from 154 farms in 5 countries. Each trait was divided into primiparous and multiparous (including second to fifth parity) groups. Genetic parameters and breeding values were estimated by using a multitrait (2-trait, 2-parity classes) repeatability model. Herd-year-season of calving, DIM, age of calving, and parity were used as fixed effects. Random effects were defined as parity (within-parity permanent environment), nongenetic cow (across-parity permanent environment), additive genetic animal, and residual effects. The estimated heritability of PNUE and PNL in the first and later parity were 0.13, 0.12, 0.14, and 0.13, and the repeatability values were 0.49, 0.40, 0.55, and 0.43, respectively. The estimated approximate genetic correlations between PNUE and PNL were negative and high (from −0.89 to −0.53), whereas the phenotypic correlations were also negative but relatively low (from −0.45 to −0.11). At a level of reliability of more than 0.30 for all novel traits, a total of 504 bulls born after 1995 had also publishable Interbull multiple-trait across-country estimated breeding values (EBV). The approximate genetic correlations between PNUE and the other 30 traits of interest, estimated as corrected correlations between EBV of bulls, ranged from −0.46 (udder depth) to 0.47 (milk yield). Obtained results showed the complex genetic relationship between efficiency, production, and other traits: for instance, more efficient cows seem to give more milk, which is linked to deeper udders, but seem to have lower health, fertility, and longevity. Additionally, the approximate genetic correlations between PNL, lower values representing less loss of N, and the 30 other traits, were from −0.32 (angularity) to 0.57 (direct calving ease). Even if further research is needed, our results provided preliminary evidence that the PNUE and PNL traits used as proxies could be included in genetic improvement programs in Holstein cows and could help their management.

The aim of this study was to identify genomic regions associated with 305‐day milk yield and lactation curve parameters on primiparous (n = 9,910) and multiparous (n = 11,158) Holstein cows. The SNP solutions were estimated using a weighted single‐step genomic BLUP approach and imputed high‐density panel (777k) genotypes. The proportion of genetic variance explained by windows of 50 consecutive SNP (with an average of 165 Kb) was calculated, and regions that accounted for more than 0.50% of the variance were used to search for candidate genes. Estimated heritabilities were 0.37, 0.34, 0.17, 0.12, 0.30 and 0.19, respectively, for 305‐day milk yield, peak yield, peak time, ramp, scale and decay for primiparous cows. Genetic correlations of 305‐day milk yield with peak yield, peak time, ramp, scale and decay in primiparous cows were 0.99, 0.63, 0.20, 0.97 and −0.52, respectively. The results identified three windows on BTA14 associated with 305‐day milk yield and the parameters of lactation curve in primi‐ and multiparous cows. Previously proposed candidate genes for milk yield supported by this work include GRINA, CYHR1, FOXH1, TONSL, PPP1R16A, ARHGAP39, MAF1, OPLAH and MROH1, whereas newly identified candidate genes are MIR2308, ZNF7, ZNF34, SLURP1, MAFA and KIFC2 (BTA14). The protein lipidation biological process term, which plays a key role in controlling protein localization and function, was identified as the most important term enriched by the identified genes.

Improving nitrogen use efficiency (NUE) at both the individual cow and the herd level has become a key target in dairy production systems, for both environmental and economic reasons. Cost-effective and large-scale phenotyping methods are required to improve NUE through genetic selection and by feeding and management strategies. The aim of this study was to evaluate the possibility of using mid-infrared (MIR) spectra of milk to predict individual dairy cow NUE during early lactation. Data were collected from 129 Holstein cows, from calving until 50 d in milk, in 3 research herds (Denmark, Ireland, and the UK). In 2 of the herds, diets were designed to challenge cows metabolically, whereas a diet reflecting local management practices was offered in the third herd. Nitrogen intake (kg/d) and nitrogen excreted in milk (kg/d) were calculated daily. Nitrogen use efficiency was calculated as the ratio between nitrogen in milk and nitrogen intake, and expressed as a percentage. Individual daily values for NUE ranged from 9.7 to 81.7%, with an average of 36.9% and standard deviation of 10.4%. Milk MIR spectra were recorded twice weekly and were standardized into a common format to avoid bias between apparatus or sampling periods. Regression models predicting NUE using milk MIR spectra were developed on 1,034 observations using partial least squares or support vector machines regression methods. The models were then evaluated through (1) a cross-validation using 10 subsets, (2) a cow validation excluding 25% of the cows to be used as a validation set, and (3) a diet validation excluding each of the diets one by one to be used as validation sets. The best statistical performances were obtained when using the support vector machines method. Inclusion of milk yield and lactation number as predictors, in combination with the spectra, also improved the calibration. In cross-validation, the best model predicted NUE with a coefficient of determination of cross-validation of 0.74 and a relative error of 14%, which is suitable to discriminate between low- and high-NUE cows. When performing the cow validation, the relative error remained at 14%, and during the diet validation the relative error ranged from 12 to 34%. In the diet validation, the models showed a lack of robustness, demonstrating difficulties in predicting NUE for diets and for samples that were not represented in the calibration data set. Hence, a need exists to integrate more data in the models to cover a maximum of variability regarding breeds, diets, lactation stages, management practices, seasons, MIR instruments, and geographic regions. Although the model needs to be validated and improved for use in routine conditions, these preliminary results showed that it was possible to obtain information on NUE through milk MIR spectra. This could potentially allow large-scale predictions to aid both further genetic and genomic studies, and the development of farm management tools.

Genetic parameters were estimated for first lactation survival defined as a binary trait (alive or dead to second calving) and the curve shape traits of milk yield, fat and protein percentages using information from 25 981 primiparous Tunisian Holsteins. For each trait, shape curves (i.e. peak lactation, persistency), level of production adjusted to 305 days in milk (DIMs) for total milk yield (TMY), and average fat (TF %) and protein (TP %) percentages were defined. Variance components were estimated with a linear random regression model under three bivariate animal models. Production traits were modelled by fixed herd × test-day (TD) interaction effects, fixed classes of 25 DIMs × age of calving × season of calving interaction effects, fixed classes of pregnancy, random environment effects and random additive genetic effects. Survival was modelled by fixed herd × year of calving interaction effects and age of calving × season of calving interaction effects, random permanent environment effects, and random additive genetic effects. Heritability (h2) estimates were 0.03 (±0.01) for survival and 0.23 (±0.01), 0.31 (±0.01) and 0.31 (±0.01) for TMY, TF % and TP %, respectively. Genetic correlations between survival and TMY, TF % and TP % were 0.26 (±0.08), −0.24 (±0.06) and −0.13 (±0.06), respectively. Genetic correlations between survival and persistency for fat and protein percentages were −0.35 (±0.09) and −0.19 (±0.09), respectively. Cows that had higher persistencies for fat and protein percentages were more likely not to survive.

International audience; Greenhouse gases (GHG) emissions from livestock and more especially methane (CH4) emissions from cattle related to ruminal fermentations remain the most important source of GHG within the agricultural sector. Main levers to reduce those emissions are the diet and the genetic selection. To study the impact of those levers and how reduce CH4 emissions, a large amount of reference measurements are needed. However, existing techniques to measure CH4 emissions from dairy cows are expensive, time consuming and difficult to apply on a large amount of animals. This is why the availability of a robust proxy to estimate individual daily CH4 emissions from dairy cows would be valuable. Estimate CH4 emissions from milk mid infrared (MIR) spectrum present potential to meet this aim as it can be obtained routinely at reasonable cost through milk recording process. Develop this equation is particularly challenging as the CH4 prediction equation from milk MIR spectra as CH4 is not a direct milk component but an indirect phenotype linked to milk composition through ruminal fermentations which theoretically influence both. To increase the variability of the calibration set, two datasets (CH4 measurements and milk MIR spectra) have been merged: A) 532 data from 156 cows of Ireland and Belgium using the SF6 tracer technique; B) 584 data from 147 cows of Switzerland, United Kingdom, France, Denmark and Germany collected in respiration chambers. In addition of the calibration using the raw reference values, a second calibration was performed with a reduction of 8 % to CH4 values from chambers evaluate the need to correct the potential method bias in accordance with literature. A 5-groups cross-validation was performed to test the robustness of the models. Those equations showed a R² and a standard error of cross-validation of 0.63 and 62 g/day respectively for the calibration on raw values and 0.65 and 59 g/day when respiration chamber values are adjusted. The slight improvement due to adjustment of chamber measurement is not significant. With errors around 60 g/day, the current equations does not permit to distinguish slight variation of CH4 emissions as it is often required in nutritional context. However, more variability is included (cows, breeds, diets, and country specific information), marked trends can be differentiated and statistics confirming its potential as proxy especially for genetic evaluations or life cycle analyses.

Recent research showed the usefulness of using estimated breeding values (EBV)for mid-infrared (MIR) based biomarkers in genetic improvement. A novel class ofbiomarkers was defined based on modelling responses of milk composition(e.g., mid-infrared (MIR) based) to stress expressed on continuous scales usingreaction norm models. Heat stress is an important aspect of dairy production even in temperate climates as shown in recent studies. Implementation of genomic selection for tolerance to heat stress is therefore not only an issue for Australian dairy cattle, a country that recently introduced such an evaluation. The question remains open if using milk composition based heat stress resilience genomically enhanced EBV (GEBV) is not a viable option. Genetic parameters were estimated for production, udder health, and milk composition traits. Data included 155,977 test-day records for milk, fat, and protein yields, fat and protein percentages, 9 individual milk fatty acids (FA), 7 FA groups, 5 minerals, and 3 ketone bodies predicted by mid-infrared spectrometry, and 7 FA groups. Data were from 21,375 first-lactation Holstein cows in 473 herds in the Walloon region of Belgium and were collected between 2008 and 2014. Test-day records were merged with daily temperature-humidity index (THI) values based on meteorological records from public weather stations. The maximumdistance between each farm and its corresponding weather station was 13 km. Linear reaction norm models were used to estimate the intercept and slope responses of 23 traits to increasing THI values. Most yield and FA traits had phenotypic and genetic declines as THI increased, whereas C18:0, C18:1 cis-9, and 4 FA groups (unsaturated FA, monounsaturated FA, polyunsaturated FA, and long-chain FA) increased with THI. Moreover, the latter traits had the largest slope-to-intercept genetic variance ratios, which indicate that they are more affected by heat stress at high THI levels, and therefore good candidate traits. Among all traits, C18:1 cis-9 was the most sensitive to heat stress. As this trait is known to reflect body reserve mobilization, using its response to THI could be a very affordable milk biomarker of heat stress for dairy cattle, expressing the equilibrium between intake and mobilization, and therefore adaptation, under warm conditions. By including novel milk based composition traits to traditional production traits, correlations between EBVs and GEBVs of those milk based traits and udder health, fertility and longevity increased considerably. Thisstudy demonstrated that milk composition resilience heat stress traits could be used as early indicators of robustness traits. Our results also suggest that marker information tend to lead to higher accuracies of prediction. Therefore, new options to improve robustness through genomic selection in Walloon Holsteins are now available.

Recent research showed the usefulness of using estimated breeding values (EBV)for mid-infrared (MIR) based biomarkers in genetic improvement. Similarly, research has also shown that genetic variation is contained in the absorbance traits along the MIR band of wavelengths. Targeted extraction of the useful genetic variance can be achieved by the combination of EBV. Direct estimation of EBV for absorbance traits was demonstrated. Our first objective was to show that the reduction of the rank of the (co)variance structure among spectral traits is possible by imposing linear functions, even if these functions represent lower accuracy MIR biomarkers. MIR based biomarkers traits were derived from ongoing research in the FP7 GplusE project. In this study, the pathway from MIR spectra to the use in genetic improvement will be described. First, blood reference phenotypic data was collected on Holstein cows, at early lactation for IGF-1, glucose, urea, cholesterol, fructosamine, beta-hydroxybutyric (BHB) acid and non-esterified fatty acids (NEFA). These traits were calibrated against corresponding MIR spectral data. Calibration Rcv 2 ranged from 0.21 to 0.51, very low from a chemometrical point of view, but potentially sufficient to extract useful spectral variation. This was validated, using EBV that were based on these MIR predictions for 144,623 records (closest to days in milk 25), from 73,378cows, in the Walloon region of Belgium. Single-trait, but multi-lactation (1, 2, 3+)models yielded h2 estimates ranging from 0.07 to 0.27. At least 20 daughters with novel traits and official EBV for udder health, fertility and longevity, with minimum reliabilities of 70% were required; a total of 124 bulls met this criteria. Standard selection index theory would usually rely on prediction error variance minimisation and estimated population (co)variances. Alternatively in this study, Partial Least Squares were applied to EBV for the milk MIR based biomarkers to develop novel genetic predictors for udder health, fertility and longevity, by extracting genetic variation along the wave band after rank reduction. Using all bulls, correlations between best predictors and EBV for udder health, fertility and longevity were at least 0.63, 0.67 and 0.62. Using selection index theory and based on significant increases of prediction abilities of longevity (0.76 compared to 0.68 from udder health or fertility alone) using milk MIR based blood biomarkers, their potential contribution to genetic improvement of udder health, fertility and longevity will be demonstrated.

Large-scale measurement of enteric methane (CH4) from individual animals is a requisite for estimation of genetic parameters and prediction of breeding values. Direct measurement of individual CH4 emissions is logistically demanding and expensive, and correlated traits (proxies) or models can be used instead as a means to predict emissions. However, most predictive models tend to be specific and are valid mainly within the circumstances under which they were developed. Robust prediction models that work across countries and production environments may be built by combining heterogeneous data from several sources. However, combining heterogeneous individual animal observations on CH4 proxies from several sources is challenging and reports are scant in literature. The main objective of this study was to combine heterogeneous individual animal observations on CH4 proxies to develop robust enteric CH4 prediction models. Data on dairy cattle CH4 emissions and related proxies from 16 herds were made available by 13 research centers across 9 European countries within the Methagene EU COST Action FA1302 consortium on “Large-scale methane measurements on individual ruminants for genetic evaluations”. After a through edition and harmonization, the final dataset comprised 48,804 observations from 2,391 cows. Random Forest (RF) models were used to predict CH4 emissions and to estimate the relative importance of proxies for across-country predictions. Principal component analysis (PCA) was used to detect potential data stratifications. Milk yield, milk fat, DIM, BW, herd and country of origin appeared to be the most relevant proxies in the prediction model. An overall prediction accuracy of 0.81 was estimated from the combined heterogeneous data. This study is a first attempt to develop methods and approaches to combine heterogeneous individual animal data on proxies for CH4 to build robust models for prediction of CH4 emissions across diverse production systems and environments. The methodology outlined here can be extended to combining heterogeneous data, pedigree information and genome-wide dense marker information for estimation of genetic parameters and prediction of breeding values for traits related to dairy system CH4 emissions. Keywords: enteric methane, heterogeneous data, prediction accuracy, methane proxies, random forest, dairy cattle., Proceedings of the World Congress on Genetics Applied to Livestock Production. Volume: Methods and Tools - Prediction 2