Using machine learning methods to predict dry matter intake from milk mid-infrared spectroscopy data on Swedish dairy cattle.

In this research communication we compare three different approaches for developing dry matter intake (DMI) prediction models based on milk mid-infrared spectra (MIRS), using data collected from a research herd over five years. In dairy production, knowledge of individual DMI could be important and useful, but DMI can be difficult and expensive to measure on most commercial farms as cows are commonly group-fed. Instead, this parameter is often estimated based on the age, body weight, stage of lactation and body condition score of the cow. Recently, milk MIRS have also been used as a tool to estimate DMI. There are different methods available to create prediction models from large datasets. The main data used were total DMI calculated as a 3-d average, coupled with milk MIRS data available fortnightly. Data on milk yield and lactation stage parameters were also available for each animal. We compared the performance of three prediction approaches: partial least-squares regression, support vector machine regression and random forest regression. The full milk MIRS alone gave low to moderate prediction accuracy (R2 = 0.07-0.40), regardless of prediction modelling approach. Adding more variables to the model improved R2 and decreased the prediction error. Overall, partial least-squares regression proved to be the best method for predicting DMI from milk MIRS data, while MIRS data together with milk yield and concentrate DMI at 3-30 d in milk provided good prediction accuracy (R2 = 0.52-0.65) regardless of the prediction tool used.

Full-lactation performance of multiparous dairy cows with differing residual feed intake.

Residual feed intake (RFI) is an efficiency trait underpinning profitability and environmental sustainability in dairy production. This study compared performance during a complete lactation of 36 multiparous dairy cows divided into three equal-sized groups with high (HRFI), intermediate (IRFI) or low RFI (LRFI). Residual feed intake was determined by two different equations. Residual feed intake according to the NorFor system was calculated as (RFINorFor) = (NEintake)-(NEmaintenance + NEgestation + NEmilk-NEmobilisation + NEdeposition). Residual feed intake according to the USA National Research Council (NRC) (RFINRC) was calculated as: RFI = DMI - predicted DMI where predicteds DMI = [(0.372× ECM)+(0.0968×BW0.75)]×(1-e-0.192×(DIM/7+3.67)). Cows in the HRFINorFor group showed higher daily CH4 production, CH4/ECM and CH4 yield (g/kg DMI) than IRFINorFor and LRFINorFor cows. Cows characterized by high efficiency (LRFINorFor) according to the NorFor system had lower body weight. Dry matter intake and apparent dry matter digestibility were not affected by efficiency group but milk yield was lower in the low efficiency, HRFINorFor, group. Cows characterized by high efficiency according to the NRC system (LRFINRC) had lower dry matter intake while yield of CH4 was higher. Daily CH4 production and CH4 g/kg ECM did not differ between RFINRC groups. Dairy cows characterized by high efficiency (both LRFINorFor and LRFINRC cows) over a complete lactation mobilized more of their body reserves in early lactation as well as during the complete lactation. The results also indicated great phenotypic variation in RFI between different stages the lactation.

In Vitro Incubations Do Not Reflect In Vivo Differences Based on Ranking of Low and High Methane Emitters in Dairy Cows.

This study evaluated if ranking dairy cows as low and high CH4 emitters using the GreenFeed system (GF) can be replicated in in vitro conditions using an automated gas system and its possible implications in terms of fermentation balance. Seven pairs of low and high emitters fed the same diet were selected on the basis of residual CH4 production, and rumen fluid taken from each pair incubated separately in the in vitro gas production system. In total, seven in vitro incubations were performed with inoculums taken from low and high CH4 emitting cows incubated in two substrates differing in forage-to-concentrate proportion, each without or with the addition of cashew nutshell liquid (CNSL) as an inhibitor of CH4 production. Except for the aimed differences in CH4 production, no statistical differences were detected among groups of low and high emitters either in in vivo animal performance or rumen fermentation profile prior to the in vitro incubations. The effect of in vivo ranking was poorly replicated in in vitro conditions after 48 h of anaerobic fermentation. Instead, the effects of diet and CNSL were more consistent. The inclusion of 50% barley in the diet (SB) increased both asymptotic gas production by 17.3% and predicted in vivo CH4 by 26.2%, when compared to 100% grass silage (S) substrate, respectively. The SB diet produced on average more propionate (+28 mmol/mol) and consequently less acetate compared to the S diet. Irrespective of CH4 emitter group, CNSL decreased predicted in vivo CH4 (26.7 vs. 11.1 mL/ g of dry matter; DM) and stoichiometric CH4 (CH4VFA; 304 vs. 235 moles/mol VFA), with these being also reflected in decreased total gas production per unit of volatile fatty acids (VFA). Microbial structure was assessed on rumen fluid sampled prior to in vitro incubation, by sequencing of the V4 region of 16S rRNA gene. Principal coordinate analysis (PCoA) on operational taxonomic unit (OTU) did not show any differences between groups. Some differences appeared of relative abundance between groups in some specific OTUs mainly related to Prevotella. Genus Methanobrevibacter represented 93.7 ± 3.33% of the archaeal sequences. There were no clear differences between groups in relative abundance of Methanobrevibacter.

Compound- and context-dependent effects of antibiotics on greenhouse gas emissions from livestock.

The use of antibiotics in livestock production may trigger ecosystem disservices, including increased emissions of greenhouse gases. To evaluate this, we conducted two separate animal experiments, administering two widely used antibiotic compounds (benzylpenicillin and tetracycline) to dairy cows over a 4- or 5-day period locally and/or systemically. We then recorded enteric methane production, total gas production from dung decomposing under aerobic versus anaerobic conditions, prokaryotic community composition in rumen and dung, and accompanying changes in nutrient intake, rumen fermentation, and digestibility resulting from antibiotic administration. The focal antibiotics had no detectable effect on gas emissions from enteric fermentation or dung in aerobic conditions, while they decreased total gas production from anaerobic dung. Microbiome-level effects of benzylpenicillin proved markedly different from those previously recorded for tetracycline in dung, and did not differ by the mode of administration (local or systemic). Antibiotic effects on gas production differed substantially between dung maintained under aerobic versus anaerobic conditions and between compounds. These findings demonstrate compound- and context-dependent impacts of antibiotics on methane emissions and underlying processes, and highlight the need for a global synthesis of data on agricultural antibiotic use before understanding their climatic impacts.

Evaluation of a gas in vitro system for predicting methane production in vivo.

Methane production from ruminant livestock varies with the diet as a result of factors such as dry matter intake, diet composition, and digestibility. To estimate the effect of dietary composition and feed additives, CH4 production can be measured in vitro as a first step because large numbers of samples can be incubated and analyzed at the same time. This study evaluated a recently developed in vitro method for prediction of in vivo CH4 production by examining the relationship between predicted and observed CH4 production values. A total of 49 different diets (observations), used in previous 13 in vivo studies, were selected to include diets varying in nutrient composition. Methane production was measured in all in vivo studies by respiration chambers or the GreenFeed system (C-Lock Inc., Rapid City, SD). Overall, the in vitro system predicted CH4 production well (R2 = 0.96), but the values obtained were slightly underestimated compared with observed in vivo values (mean 399 L/d compared with 418 L/d: root mean square prediction error = 51.6 L/d or 12.3% of observed mean). Further analysis of the effect on residuals showed no significant relationship between CH4 production and most factors known to affect CH4 production such as dry matter intake, digestibility, and dietary concentrations of fat and starch. However, some factors included in the model were not well predicted by the system, with residuals negatively related to neutral detergent fiber concentration and positively related to concentrate proportion. The in vitro system can thus be useful for screening diets and evaluation of feed additives as a first step that can be best interpreted when feeding cows at maintenance level.

Methane Production in Dairy Cows Correlates with Rumen Methanogenic and Bacterial Community Structure.

Methane (CH4) is produced as an end product from feed fermentation in the rumen. Yield of CH4 varies between individuals despite identical feeding conditions. To get a better understanding of factors behind the individual variation, 73 dairy cows given the same feed but differing in CH4 emissions were investigated with focus on fiber digestion, fermentation end products and bacterial and archaeal composition. In total 21 cows (12 Holstein, 9 Swedish Red) identified as persistent low, medium or high CH4 emitters over a 3 month period were furthermore chosen for analysis of microbial community structure in rumen fluid. This was assessed by sequencing the V4 region of 16S rRNA gene and by quantitative qPCR of targeted Methanobrevibacter groups. The results showed a positive correlation between low CH4 emitters and higher abundance of Methanobrevibacter ruminantium clade. Principal coordinate analysis (PCoA) on operational taxonomic unit (OTU) level of bacteria showed two distinct clusters (P < 0.01) that were related to CH4 production. One cluster was associated with low CH4 production (referred to as cluster L) whereas the other cluster was associated with high CH4 production (cluster H) and the medium emitters occurred in both clusters. The differences between clusters were primarily linked to differential abundances of certain OTUs belonging to Prevotella. Moreover, several OTUs belonging to the family Succinivibrionaceae were dominant in samples belonging to cluster L. Fermentation pattern of volatile fatty acids showed that proportion of propionate was higher in cluster L, while proportion of butyrate was higher in cluster H. No difference was found in milk production or organic matter digestibility between cows. Cows in cluster L had lower CH4/kg energy corrected milk (ECM) compared to cows in cluster H, 8.3 compared to 9.7 g CH4/kg ECM, showing that low CH4 cows utilized the feed more efficient for milk production which might indicate a more efficient microbial population or host genetic differences that is reflected in bacterial and archaeal (or methanogens) populations.

Effects on enteric methane production and bacterial and archaeal communities by the addition of cashew nut shell extract or glycerol-an in vitro evaluation.

The objective of the study was to evaluate the effect of cashew nut shell extract (CNSE) and glycerol (purity >99%) on enteric methane (CH4) production and microbial communities in an automated gas in vitro system. Microbial communities from the in vitro system were compared with samples from the donor cows, in vivo. Inoculated rumen fluid was mixed with a diet with a 60:40 forage:concentrate ratio and, in total, 5 different treatments were set up: 5mg of CNSE (CNSE-L), 10mg of CNSE (CNSE-H), 15mmol of glycerol/L (glycerol-L), and 30mmol of glycerol/L (glycerol-H), and a control without feed additive. Gas samples were taken at 2, 4, 8, 24, 32, and 48h of incubation, and the CH4 concentration was measured. Samples of rumen fluid were taken for volatile fatty acid analysis and for microbial sequence analyses after 8, 24, and 48h of incubation. In vivo rumen samples from the cows were taken 2h after the morning feeding at 3 consecutive days to compare the in vitro system with in vivo conditions. The gas data and data from microbial sequence analysis (454 sequencing) were analyzed using a mixed model and principal components analysis. These analyses illustrated that CH4 production was reduced with the CNSE treatment, by 8 and 18%, respectively, for the L and H concentration. Glycerol instead increased CH4 production by 8 and 12%, respectively, for the L and H concentration. The inhibition with CNSE could be due to the observed shift in bacterial population, possibly resulting in decreased production of hydrogen or formate, the methanogenic substrates. Alternatively the response could be explained by a shift in the methanogenic community. In the glycerol treatments, no main differences in bacterial or archaeal population were detected compared with the in vivo control. Thus, the increase in CH4 production may be explained by the increase in substrate in the in vitro system. The reduced CH4 production in vitro with CNSE suggests that CNSE can be a promising inhibitor of CH4 formation in the rumen of dairy cows.