Previous Article in Journal
Fatty Acid Composition and Lipid Oxidation in Plant-Based Meat Analogue Chicken Schnitzels Under Different Cooking Conditions
 
 
Search for Articles:
Title / Keyword
Author / Affiliation / Email
Journal
Article Type
 
 
Advanced Search
 
Section
Special Issue
Volume
Issue
Number
Page
 
Journals
Lipidology
Volume 2
Issue 4
10.3390/lipidology2040024
lipidology-logo
Submit to this Journal Review for this Journal Propose a Special Issue
► Article Menu

Article Menu

share Share announcement Help format_quote Cite question_answer Discuss in SciProfiles
first_page
settings
Order Article Reprints
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

Milk Fatty Acid Profiling as a Tool for Estimating Methane Emissions in Conventionally Fed Dairy Cows

by
Emily C. Youngmark
1 and
Jana Kraft
1,2,3,*
1
Department of Animal and Veterinary Sciences, The University of Vermont, Burlington, VT 05405, USA
2
Department of Medicine, Division of Endocrinology, Metabolism and Diabetes, The University of Vermont, Colchester, VT 05446, USA
3
Department of Nutrition and Food Sciences, The University of Vermont, Burlington, VT 05405, USA
*
Author to whom correspondence should be addressed.
Lipidology 2025, 2(4), 24; https://doi.org/10.3390/lipidology2040024
Submission received: 14 October 2025 / Revised: 17 November 2025 / Accepted: 1 December 2025 / Published: 2 December 2025
Browse Figures
Review Reports Versions Notes

Abstract

Milk fatty acid (FA) synthesis and enteric methanogenesis share common biochemical pathways related to rumen fermentation patterns and microbial volatile FA production. The FA profile of milk is known to correlate with methane (CH4) emissions; thus, FA profiling has been proposed as an indirect method to predict CH4 emissions from dairy cattle. This study aimed to (1) investigate the milk FA profiles of Holstein cows to identify candidate biomarkers for predicting CH4 output (g/d), CH4 yield (g/kg dry matter intake), and CH4 intensity (g/kg energy-corrected milk), and (2) develop and compare regression models predicting CH4 emissions. Forty-eight cows, fed industry standard diets, were enrolled in an exploratory trial. Milk samples and CH4 measurements were collected thrice per day, and intake was recorded daily. Milk lipids were extracted, transesterified, and subsequently analyzed via gas–liquid chromatography. Three penalized regression models were compared for predicting CH4 emission metrics using milk FAs and management variables. Methane emission metrics corelated positively with short- and medium-chain FAs, polyunsaturated FAs, and branched-chain FAs, while monounsaturated FAs correlated negatively. Notably, this study observed novel correlations between 11-cyclohexyl-11:0; and 20:3 c5,c8,c11 and CH4 metrics (|r| = 0.58–0.79). Across all CH4 metrics, the models demonstrated high predictive accuracy (R2 = 0.71–0.87; concordance correlation coefficient = 0.83–0.93). The findings of this study indicate that milk FA profiling may be an effective method to detect CH4 emissions from cows fed industry standard diets and highlight the need for further refinement of prediction models.

Graphical Abstract

1. Introduction

Methane (CH4) mitigation on dairy farms is necessary to reduce dairy production’s impact on climate change. Enteric fermentation from ruminant livestock is responsible for approximately 25% of U.S. anthropogenic CH4 emissions [1]. Widespread adoption of CH4-reducing sustainable practices on commercial farms requires the development of CH4 measurement methods that are easy to use. Many techniques are available to measure CH4 emissions for research purposes (reviewed in Youngmark and Kraft [2]). These, however, can be difficult to incorporate into established production systems as they are generally expensive, may alter the daily routine of the animals and laborers, and require specialized equipment and labor (reviewed in Youngmark and Kraft [2] and Hammond et al. [3]).
Milk fatty acid (FA) profiling may bridge the gap between research and on-farm CH4 measurement. The composition of milk FAs is greatly influenced by diet and rumen fermentation patterns (reviewed in Lock and Bauman [4] and Dewhurst et al. [5]), both of which drive methanogenesis [6]. De novo milk FA synthesis is reliant on mammary gland uptake of acetate and β-hydroxybutyrate (BHBA), two volatile FAs (VFAs) that are positively associated with CH4 emissions [7,8]. Consistently, milk FAs derived from de novo synthesis [i.e., short- and medium-chain FAs (SMCFAs)] tend to positively correlate with CH4 emissions [9,10,11]. In contrast, propionate synthesis is negatively correlated with CH4 emissions [8]. Propionate is the primary precursor for gluconeogenesis, mediates milk yield, and is a substrate for odd-chain FA (OCFA) synthesis by rumen microbes [6]. Monounsaturated FAs (MUFAs) and polyunsaturated FAs (PUFAs) found in milk are predominantly derived from the diet or arise as biohydrogenation intermediates of dietary PUFAs (reviewed in Taormina et al. [12]). These FAs are generally reported to be negatively correlated with CH4 emissions [12,13].
Previous studies have identified many individual and groupings of milk FAs that correlate with CH4 emissions. However, the strength and direction of correlations between emissions and specific milk FAs differ between studies, and very few FAs are consistently included in prediction models across studies [7,14]. These discrepancies are typically attributed to dietary differences, as the experimental diets used in each study vary significantly [7,9,15]. In addition, the experimental diets used in previous studies are not representative of industry standards, often including high levels of fat, concentrates, and/or anti-methanogenic compounds [10,11]. These dietary manipulations significantly alter rumen fermentation, biohydrogenation, microbial communities, and microbial metabolites (reviewed in Toral et al. [16]), all of which contribute to the milk FA profile and rumen methanogenesis. Therefore, prediction models derived from studies using such experimental diets may not fully capture the natural variation in CH4 emissions or the milk FA profile of cows in standard (i.e., conventional) production systems. To our knowledge, no studies have directly examined the relationship between CH4 emissions and the milk FA profiles of cows fed solely standard production-based diets.
We hypothesized that milk FA profiling can be used as a method to predict inherent differences in CH4 emissions from cows fed industry standard diets. The objectives of this study were to (1) analyze the milk FA profiles across a lactating herd fed industry standard diets to identify FA candidates that may serve as biomarkers for CH4 emissions, and (2) construct and evaluate ridge, LASSO, and elastic net regression models to predict CH4 emission metrics using milk FA profiles and production parameters as explanatory variables.

2. Materials and Methods

2.1. Animals, Experimental Design, and Diets

All experimental protocols involving cows were approved by the University of Vermont’s Institutional Animal Care and Use Committee (Protocol#: IPROTO202300000008). A total of 48 lactating Holstein cows (12 primiparous, 36 multiparous) were enrolled into an exploratory study in the Paul Miller Research Complex in April of 2023. Of these, 8 cows were in early lactation (≤100 days in milk; DIM), 18 cows were in mid-lactation (101 ≤ DIM ≤ 200), and 22 cows were in late lactation (≥201 DIM). Cows were moved to tie-stalls 12 h prior to the 48 h observation period. The average barn temperature across the trial was 14.5 ± 3.7 °C. The cows were fed a total mixed ration (TMR) twice daily at 0700 h and 1200 h. The TMR was formulated by Poulin Grain (Newport, VT, USA) for estimated milk production and parity (high-yield TMR: ≥45 kg milk/d, n = 24; low-yield TMR: <45 kg milk/d, n = 12; and a primiparous TMR: 40 kg milk/d, n = 12; Table 1). Feed and water were provided ad libitum. Cows were milked thrice daily at 0330, 1130, and 1930 h.

2.2. Intake and Diet Composition

Feed intake and refusals were recorded, and representative samples were collected daily for dry matter intake (DMI) calculation and composition analysis. Samples were stored at −20 °C until analysis. The chemical composition [i.e., neutral detergent fiber (NDF), acid detergent fiber (ADF), crude fiber, crude protein (CP), starch, ether extract (EE), and sugars] of diets (Table 2) were determined via wet chemistry analysis (Dairy One, Ithaca, NY, USA).

2.3. Milk Yield and Composition

Milk yield was recorded electronically at each milking. Milk samples were collected from individual cows through an in-line milk sampling system (Allflex, Madison, WI, USA) at each milking. The in-line milk sampling system continuously collects proportional aliquots throughout the entire milking event, ensuring a representative composite sample that reflects the cow’s total milk output. Approximately 45 mL were collected, inoculated with a preservative (Broad Spectrum Microtab II; Advanced Instruments, LLC, Norwood, MA, USA) and stored at 4 °C before being shipped to the William H. Miner Agricultural Research Institute (Chazy, NY, USA) for composition analysis (i.e., fat, anhydrous lactose, true protein, total solids, nonfat solids, urea nitrogen, and somatic cell count). A second aliquot (50 mL) was collected and stored at −20 °C for FA analysis. Samples for each cow were composited by day in proportion to the milk yield and fat percentage from each milking. Composited samples were centrifuged at 17,800× g (Sorvall RC-5B; Du Point, Wilmington, DE, USA) for 30 min at 8 °C. The cream layer was transferred to a collection tube and stored at −20 °C until further processing.

2.4. Methane Measurement

Gas emissions were spot-sampled thrice daily following each milking (0400, 1200, and 2000 h) using the GreenFeed tie-stall system (C-Lock, Rapid City, SD, USA). Cows were trained to use the GreenFeed 14 d before the beginning of the trial. During the collection period, feed was pushed to the side immediately before the GreenFeed system was placed in front of the cow. GreenFeed identified cows using a radio-frequency identification scanner (RFID). Roughly 35 g of bait feed (Textra 16%; Poulin Grain, Newport, VT, USA) was dropped into the collection tray every 30 s (10 drops allowed per visit) while an RFID tag was in range. The cow would receive approximately 355 g of bait feed during each use while gas concentrations (CH4, CO2, H2, and O2) were recorded. Raw gas measurements were analyzed by C-Lock. Standard calibration was performed daily during the trial and CO2 calibration was performed prior to the trial.

2.5. Fatty Acid Analysis

Total lipid extraction was performed using hexane:isopropanol (3:2) as described by Hara and Radin [17]. Fatty acid methyl esters (FAMEs) were prepared and identified as described in Bainbridge et al. [18] and Unger et al. [19]. Briefly, a total of 110 FAs were identified with a chain length of 4:0–24:0. The analysis was performed on a GC-2010 gas chromatograph (Shimadzu, Kyoto, Japan) equipped with an AOC-20s autosampler, an AOC-20i autoinjector, a split/splitless injector (injection volume of 1 µL; 100:1 split ratio), an SP-2560 fused-silica capillary column (100 m × 0.25 mm i.d. × 0.2 μm film thickness; Supelco Inc., Bellefonte, PA, USA), and a flame-ionization detector (FID). Fatty acids were determined via comparison of retention times with known FAME standards (Nu-Check Prep standards #463 and #674 [Elysian, MN, USA]; Supelco PUFA-3 mixture and linoleic and linolenic acid mixture [Bellefonte, PA, USA]; Larodan Fine Chemicals iso and anteiso BCFA isomers [Malmö, Sweden]; and a house-made milk standard). Short-chain FAMEs were corrected for mass discrepancy as per Wolff et al. [20]. The identification of FAMEs (except 11-cyclohexyl-11:0) was verified in-house using GC-MS [21]. The verification of 11-cyclohexyl-11:0 was based on the spectral confirmation reported by Shingfield et al. [22].

2.6. Statistical Analysis

Before statistical analysis, all data were averaged over the 48 h sampling period. Cows with less than two GreenFeed observations per day were not considered for analysis (n = 8). Spearman’s correlation analyses between milk FAs (% of total FAMEs identified) and CH4 output (g/d), CH4 yield (g/kg DMI), and CH4 intensity (g/kg ECM) were calculated on individual values (n = 40) using the cor() procedure from the stats package in R Studio, and probability values were determined using the corr.test() procedure from the psych package in R Studio. Correlations were defined as weak (|r| < 0.4), moderate (0.4 ≤ |r| < 0.6), and strong (|r| ≥ 0.6).

2.6.1. Variable Selection for Regression Analysis

Variables with an absolute correlation to the response of |r| ≤ 0.5 were retained for multivariate modeling. Low-abundance milk FA (< 0.05% FAME; i.e., 5:0; 7:0; 9:0; 13:0-iso; 14:1 t9; 16:1 t9; 16:1 c10/t13; 17:1 c7; 17:1 t10; 18:1 t4; 18:1 t5; 18:2 t10,t14; 18:2 c12,t16; 18:3 c6,c9,c12; 20:1 c11; 20:2 c11,c14; 20:5 c5,c8,c11,c14,c17; 22:0; 22:4 c7,c10,c13,c16; 22:5 c7,c10,c13,c16,c19; and 24:0) and unknown 16:1 isomers were not included in the regression analysis. Candidate predictors were entered into a stepwise linear regression using the step() function in R, which applies bidirectional variable selection based on the Akaike Information Criterion (AIC). This exploratory step was used to identify a smaller subset of predictors to be tested in the penalized regression framework.

2.6.2. Penalized Regression Models

To account for residual multicollinearity and overfitting, three penalized regression techniques were implemented in R Studio (R version 4.4.0) using the cv.glmnet() procedure from the glmnet package: ridge regression (α = 0), least absolute shrinkage and selection operator (LASSO) regression (α = 1), and elastic net (α = 0.5). Ridge regression retained all correlated predictors, prioritizing shrinkage and multicollinearity control, whereas LASSO and elastic net selectively shrank or zeroed coefficients [23]. Each model was fit using 10-fold cross-validation to identify the optimal regularization parameter (λ) that minimized mean cross-validated error (lambda.min). Residual plots, Q–Q plots, and coefficient path visualizations for each penalty type were visually assessed for heteroscedasticity and normality.
Model performance was evaluated using multiple goodness-of-fit and predictive accuracy statistics, as described by Bougouin et al. [14], including the coefficient of determination (R2), mean square error (MSE), root mean square error (RMSE), root mean square prediction error (RMSPE), and Lin’s concordance correlation coefficient (CCC). To assess coefficient robustness, bootstrap resampling (n = 200 iterations) was applied to each penalized model at its optimal λ value. The standard deviation of bootstrapped coefficients was used to compute a relative variance (RV) index (SD/|β|) for each predictor, which was classified as stable (RV < 0.25), moderate (0.25 ≤ RV ≤ 0.75), or unstable (RV > 0.75). This allowed for the comparison of model sensitivity to sample variation and the identification of reliable predictors.

3. Results

3.1. Intake, Performance, and Milk Composition

The mean DMI was 25.9 ± 2.2 kg/d (Table 3). The average milk yield and ECM were 40.6 ± 8.4 kg/d and 41.3 ± 8.5 kg/d, respectively, while milk fat and protein yields averaged 1.7 ± 0.41 and 1.3 ± 0.2 kg/d, respectively (Table 3). Days in milk ranged from 28 to 317, with the average cow in mid-lactation (Table 3). Average CH4 emissions were 366 ± 72 g/d (range: 238 to 516 g/d), CH4 yield averaged 14.1 ± 2.3 g/kg DMI, and CH4 intensity averaged 9.4 ± 3.3 g/kg ECM (Table 3). Descriptive statistics for the milk FA profile are summarized in Table 4.

3.2. Milk FA Correlations with CH4 Emission Metrics

3.2.1. Positive Correlations

Total SFAs and SMCFAs moderately correlated with CH4 output (r = 0.47–0.58; p ≤ 0.002) and exhibited weaker correlations with yield and intensity (r = 0.31–0.45; p < 0.04; Table 5, Figure 1). Consistent with these findings, 6:0, 8:0, 10:0, 12:0, and 14:0 moderately correlated with CH4 output (r = 0.40–0.51; p ≤ 0.01); 12:0 and 14:0 moderately correlated with CH4 yield (r = 0.40; p = 0.01); and 14:0 moderately correlated with intensity (r = 0.46; p > 0.003; Table 5, Figure 1). Total odd- and branched-chain FAs (OBCFAs), total BCFAs, as well as total iso- and anteiso-BCFAs exhibited moderate-to-strong correlations with all CH4 emission metrics (r = 0.47–0.72; p ≤ 0.002; Table 5, Figure 1). In agreement, the BCFAs 13:0-anteiso, 14:0-iso, 15:0-iso, 15:0-anteiso, 16:0-iso, and 11-cyclochexy-11:0 showed moderate-to-strong correlations with all CH4 emission metrics (r = 0.51–0.75; p ≤ 0.001 Table 5, Figure 1). Additionally, the sum of all FAs with a chain length of 20 or greater (i.e., very-long-chain FAs; VLCFAs) was strongly correlated with all emission metrics (r = 0.65–0.79), with the individual VLCFAs such as 20:1 c9; 20:3 c5,c8,c11; and 20:4 c5,c8,c11,c14 exhibiting moderate-to-strong correlations (r = 0.41–0.68; p ≤ 0.01; Table 5, Figure 1).

3.2.2. Negative Correlations

Total MUFAs, total 18:1, and total trans- and cis-18:1 isomers moderately negatively correlated with CH4 output (r = −0.46 to −0.51; p ≤ 0.003), but only weakly with CH4 yield and intensity (r ≥ −0.30; p ≤ 0.06; Table 5, Figure 1). Individual MUFAs, 17:1 c9, 18:1 t10, and 18:1 c13, were weakly to moderately negatively correlated with CH4 emission metrics (r = −0.38 to −0.54; p ≤ 0.02), while 16:1 c8 and 18:1 c11 exhibited moderate-to-strong negative correlations (r = −0.50 to −0.77; p ≤ 0.001; Table 5, Figure 1).

3.3. Regression Models

The model performance summaries for CH4 emissions are outlined in Table 6. The corresponding variables and standardized coefficients for each model are given in Table 7. Methane output models showed high prediction accuracy with low prediction error (R2 = 0.79–0.85; CCC = 0.87–0.92; RMSPE ≤ 0.09) and RMSE of 27.6–32.7 g/d (Table 6). Methane yield models had the lowest prediction accuracy but still performed reasonably well (R2 = 0.71–0.78; CCC = 0.82–0.87; RMSPE ≤ 0.09; Table 6). The RMSE range for CH4 yield models was between 1.05 and 1.20 g/kg DMI (Table 6). Methane intensity models showed the strongest overall performance (R2 = 0.80–0.87; CCC = 0.88–0.93), with a RMSE range of 1.15–1.45 (Table 6). These models, however, had the highest relative error (RMSPE = 0.14–0.15; Table 6).
Across prediction models, several milk FAs emerged as robust predictors of CH4 emission metrics, regardless of the modeling approach. The BCFAs 11-cyclohexyl-11:0, 13:0-anteiso, 15:0-iso, and 16:0-iso were the most common positive predictors of CH4 output, yield, and intensity (Table 7). The second-most-common positive predictors in the models were the individual VLCFAs 20:3 c5,c8,c11 and 20:4 c5,c8,c11,c14, as well as the sum of all VLCFAs (Table 7). In contrast, total SMCFAs were negatively associated with CH4 yield and intensity but were not included as a predictor of CH4 output (Table 7). Milk yield was negatively associated with both CH4 output and yield, while NDF intake was positively related to CH4 output (Table 7). Categorical variables, such as diet type and lactation stage, affected predictions but showed lower stability, suggesting that these effects were secondary to milk FA profiles and intake variables (Table 7).

4. Discussion

This study aimed to identify milk FAs that could serve as biomarkers for enteric CH4 emissions and to develop regression models predicting CH4 emissions using candidate milk FAs and production parameters. Our findings revealed moderate-to-strong relationships between milk FAs and CH4 emission metrics, highlighting the potential of milk FAs as emission biomarkers. While the observed correlations between CH4 parameters and milk FAs do not imply a cause-and-effect relationship, consistent correlations between these variables across studies suggest biological relevance. The correlations identified in this study were generally consistent with previous reports linking milk FA profiles to CH4 emissions. Positive associations between CH4 emission metrics and total SFAs, total SMCFAs, total BCFAs, and many individual FAs within these classes align with the findings of Chilliard et al. [10], Castro-Montoya et al. [9], and Bougouin et al. [14]. These results are biologically plausible given that FAs in these categories are closely tied to ruminal acetate and butyrate production, key end products of cellulolytic fermentation that are strongly associated with methanogenesis [24,25]. Similarly, the negative correlations between individual and total MUFAs, particularly those of the 17:1 c9 and 18:1 isomers, are consistent with previous reports [7,26,27]. These MUFAs are derived either directly from the diet (i.e., 18:1 c9) or are products of the incomplete biohydrogenation of dietary PUFAs (primarily 18:2 c9,c12 and 18:3 c9,c12,c15), both of which are known to inhibit CH4 emissions [27].
The inclusion of 16:0-iso and 18:1 c11, as well as intake and production parameters, in the present prediction models is largely consistent with previous studies [7,14,27], underscoring the value of these variables as predictors of enteric CH4 emissions across diets. Conversely, several of the FAs incorporated in our models are rarely included, or absent, in existing prediction equations. To the best of our knowledge, 13:0-anteiso and 20:1 c9 are reported only in equations developed by Mohammed et al. [11], 15:0-iso only in van Gastelen et al. [15], and 14:0 solely in simple linear equations by Castro-Montoya et al. [9]. Notably, 11-cyclohexyl-11:0; 20:3 c5,c8,c11; and 20:4 c5,c8,c11,c14 have not been previously incorporated in any CH4 prediction equations. The emergence of these FAs in our prediction models, together with their limited occurrence in the previous literature, highlights the need for continued research. Because they are under-represented, it remains unclear if these FAs serve as diet-specific predictors or act as potential biomarkers across diets.
Importantly, two unique milk FAs, the BCFA 11-cyclohexyl-11:0 (ω-cyclohexylundecanoic acid) and the n-9 FA 20:3 c5,c8,c11 (mead acid), emerged as strong predictors of CH4 emissions. While both FAs are known to occur in bovine milk [28,29], to our knowledge, neither milk FA has been reported in CH4-related literature thus far. Little is known about the origin of 11-cyclohexyl-11:0, though it is associated with acidophilic microbes [28,30]. This association could explain the positive correlation with CH4 parameters in the present study. The physiological source of mead acid in milk is less clear. This FA is synthesized predominantly in the liver during essential fatty acid deficiency [31,32] and functions as a lipid mediator under oxidative stress [33,34]. In goats, a greater content of mead acid in milk has been linked with greater circulating BHBA concentrations [34], suggesting altered energy metabolism. Although blood metabolites were not measured in the present study, a similar mechanism could underlie the positive correlation between mead acid and CH4 emission metrics, potentially reflecting shifts in hepatic lipid mobilization and/or oxidative status associated with rumen fermentation efficiency.
The predictive performance of the penalized regression models utilized in the present study compares favorably with those reported in the previous literature. van Lingen et al. [27] reported model R2 values of 0.47–0.75 using milk FAs and metabolites in mixed linear regression models. Similarly, Rico et al. [26] reported high R2 and low RMSE values (R2 = 0.80–0.84; RMSE = 23.52–26.03 g CH4/d) when milk FAs and intake parameters were included in LASSO and SARS regression models. In addition, the best models reported in Engelke et al. [7] and Bougouin et al. [14] included milk FA variables, intake parameters, and animal factors using linear regression and linear mixed-effects models, respectively. Engleke et al. [7] observed R2 values of 0.61–0.91, CCCs of 0.75–0.95, and RMSEs of 38.8–77.8 L CH4/d, while Bougouin et al. [14] observed CCCs of 0.73–0.84, RMSPEs of 0.14–0.17, and RMSEs of 46.6 g CH4/d, 2.6 g CH4/kg DMI, and 2.7 g CH4/kg milk. In comparison, our models exhibited comparable accuracy (R2 = 0.78–0.87; CCC = 0.86–0.91) but lower prediction error (RMSPE = 0.08–0.16; RMSE = 27.5–32.7 g CH4/d, 1.05–1.20 g CH4/kg DMI, and 1.15–1.45 g CH4/kg ECM) when integrating a range of milk FAs, intake parameters, and management factors. These results demonstrate that penalized regression methods such as ridge, LASSO, and elastic net provide robust and reliable models for predicting CH4 emission metrics from milk FA and production variables.
While the observed relationships between milk FAs and CH4 emissions are consistent with the previous literature and established rumen microbial metabolic pathways, this study cannot empirically confirm the underlying mechanisms as rumen fluid was not collected. Furthermore, the short observation period may limit the generalizability of these findings to dairy production systems, as methane emissions are known to vary within cows across lactation [35,36]. Finally, including both multi- and primiparous cows at all stages of lactation was as much a strength of this research as it was a limitation. The purpose of this inclusion criterion was to determine biomarkers indicative of all cows, regardless of parity or DIM. This variability in lactation stage and parity broadens the applicability of our research by reflecting a wider range of dairy herd conditions, though it also could have introduced confounding factors that masked relationships between milk FAs and CH4 emissions. Despite these limitations, the present study’s findings demonstrate the utility of milk FAs as biomarkers for predicting CH4 emissions and provide valuable insight into the predictive capabilities of underexplored FA. Future work should validate these models across diverse diets and production systems and explore the metabolic pathways linking specific FAs to CH4 synthesis.

5. Conclusions

Our study demonstrated distinct relationships between milk FA composition and CH4 emission metrics, indicating that FA profiling can serve as a reliable foundation to estimate emissions from cows fed industry standard diets. The inclusion of both well-established and underexplored FA predictors in our models suggests that some FAs may be indicative of CH4 emissions regardless of diet, while others may be diet-specific. This work furthers efforts to provide alternative methods for methane estimation from the dairy industry, though continued refinement of prediction models and the use of more representative diets in research is still needed to improve the applicability of FA profiling to industry settings.

Author Contributions

Conceptualization, E.C.Y. and J.K.; methodology, E.C.Y. and J.K.; validation, E.C.Y. and J.K.; formal analysis, E.C.Y.; investigation, E.C.Y. and J.K.; resources, J.K.; data curation, E.C.Y. and J.K.; writing—original draft preparation, E.C.Y.; writing—review and editing, J.K.; visualization, E.C.Y.; supervision, J.K.; project administration, J.K.; funding acquisition, J.K. All authors have read and agreed to the published version of the manuscript.

Funding

This research was funded by USDA Hatch Fund: VT-H02902.

Institutional Review Board Statement

This study was conducted in accordance with the University of Vermont’s Institutional Animal Care and Use Committee (Protocol# IPROTO202300000008; approved on 23 March 2023).

Informed Consent Statement

Not Applicable.

Data Availability Statement

Data is available upon reasonable request.

Acknowledgments

The authors want to thank the members of the Greenwood lab for their assistance with sample collection, methane measurement, and feed management during the trial.

Conflicts of Interest

The authors declare no conflicts of interest.

Abbreviations

The following abbreviations are used in this manuscript:
ADFAcid detergent fiber
AICAkaike information criterion
aNDFom/NDFAsh-free neutral detergent fiber organic matter
BCFABranched-chain fatty acid
BCSBody condition score
BHBABeta-hydroxybutyrate
CCCConcordance correlation coefficient
CH4Methane
CO2Carbon dioxide
DIMDays in milk
DMDry matter
DMIDry matter intake
ECMEnergy-corrected milk
FAFatty acid
FAMEFatty acid methyl ester
FIDFlame-ionization detector
H2Dihydrogen
MUFAMonounsaturated fatty acid
MSEMean square error
NEGNet energy for growth
NELNet energy for lactation
NEMNet energy for maintenance
NFCNon-fiber carbohydrate
OBCFAOdd- and branched-chain fatty acid
OCFAOdd-chain fatty acid
PUFAPolyunsaturated fatty acid
RFIDRadio-frequency identification
RMSERoot mean square error
RMSPERoot mean square prediction error
RVRelative variance
SDStandard deviation
SFASaturated fatty acid
SMCFAShort- and medium-chain fatty acid
TMRtotal mixed ration
VFAVolatile fatty acid
VLCFAVery-long-chain fatty acid

References

  1. EPA. Overview of Greenhouse Gases. Available online: https://www.epa.gov/ghgemissions/overview-greenhouse-gases (accessed on 4 December 2024).
  2. Youngmark, E.C.; Kraft, J. Milk fatty acids as potential biomarkers of enteric methane emissions in dairy cattle: A review. Animals 2025, 15, 2212. [Google Scholar] [CrossRef]
  3. Hammond, K.J.; Crompton, L.A.; Bannink, A.; Dijkstra, J.; Yáñez-Ruiz, D.R.; O’Kiely, P.; Kebreab, E.; Eugène, M.A.; Yu, Z.; Shingfield, K.J.; et al. Review of current in vivo measurement techniques for quantifying enteric methane emission from ruminants. Anim. Feed Sci. Technol. 2016, 219, 13–30. [Google Scholar] [CrossRef]
  4. Lock, A.L.; Bauman, D.E. Modifying milk fat composition of dairy cows to enhance fatty acids beneficial to human health. Lipids 2004, 39, 1197–1206. [Google Scholar] [CrossRef] [PubMed]
  5. Dewhurst, R.J.; Shingfield, K.J.; Lee, M.R.F.; Scollan, N.D. Increasing the concentrations of beneficial polyunsaturated fatty acids in milk produced by dairy cows in high-forage systems. Anim. Feed Sci. Technol. 2006, 131, 168–206. [Google Scholar] [CrossRef]
  6. Vlaeminck, B.; Fievez, V.; Tamminga, S.; Dewhurst, R.J.; van Vuuren, A.; De Brabander, D.; Demeyer, D. Milk odd- and branched-chain fatty acids in relation to the rumen fermentation pattern. J. Dairy Sci. 2006, 89, 3954–3964. [Google Scholar] [CrossRef]
  7. Engelke, S.W.; Das, G.; Derno, M.; Tuchscherer, A.; Wimmers, K.; Rychlik, M.; Kienberger, H.; Berg, W.; Kuhla, B.; Metges, C.C. Methane prediction based on individual or groups of milk fatty acids for dairy cows fed rations with or without linseed. J. Dairy Sci. 2019, 102, 1788–1802. [Google Scholar] [CrossRef]
  8. Ungerfeld, E.M. Metabolic hydrogen flows in rumen fermentation: Principles and possibilities of interventions. Front. Microbiol. 2020, 11, 589. [Google Scholar] [CrossRef]
  9. Castro-Montoya, J.; Campeneere, S.D.; Baets, B.D.; Fievez, V. The potential of milk fatty acids as biomarkers for methane emissions in dairy cows: A quantitative multi-study survey of literature data. J. Agric. Sci. 2016, 154, 515–531. [Google Scholar] [CrossRef]
  10. Chilliard, Y.; Martin, C.; Rouel, J.; Doreau, M. Milk fatty acids in dairy cows fed whole crude linseed, extruded linseed, or linseed oil, and their relationship with methane output. J. Dairy Sci. 2009, 92, 5199–5211. [Google Scholar] [CrossRef]
  11. Mohammed, R.; McGinn, S.M.; Beauchemin, K.A. Prediction of enteric methane output from milk fatty acid concentrations and rumen fermentation parameters in dairy cows fed sunflower, flax, or canola seeds. J. Dairy Sci. 2011, 94, 6057–6068. [Google Scholar] [CrossRef]
  12. Taormina, V.M.; Unger, A.L.; Kraft, J. Full-fat dairy products and cardiometabolic health outcomes: Does the dairy-fat matrix matter? Front. Nutr. 2024, 11, 1386257. [Google Scholar] [CrossRef]
  13. Angulo, J.; Hiller, B.; Olivera, M.; Mahecha, L.; Dannenberger, D.; Nuernberg, G.; Losand, B.; Nuernberg, K. Dietary fatty acid intervention of lactating cows simultaneously affects lipid profiles of meat and milk. J. Sci. Food Agric. 2012, 92, 2968–2974. [Google Scholar] [CrossRef]
  14. Bougouin, A.; Appuhamy, J.A.D.R.N.; Ferlay, A.; Kebreab, E.; Martin, C.; Moate, P.J.; Benchaar, C.; Lund, P.; Eugène, M. Individual milk fatty acids are potential predictors of enteric methane emissions from dairy cows fed a wide range of diets: Approach by meta-analysis. J. Dairy Sci. 2019, 102, 10616–10631. [Google Scholar] [CrossRef] [PubMed]
  15. van Gastelen, S.; Antunes-Fernandes, E.C.; Hettinga, K.A.; Dijkstra, J. Relationships between methane emission of Holstein Friesian dairy cows and fatty acids, volatile metabolites and non-volatile metabolites in milk. Animal 2017, 11, 1539–1548. [Google Scholar] [CrossRef] [PubMed]
  16. Toral, P.G.; Hervás, G.; Frutos, P. Invited review: Research on ruminal biohydrogenation—Achievements, gaps in knowledge, and future approaches from the perspective of dairy science. J. Dairy Sci. 2024, 107, 10115–10140. [Google Scholar] [CrossRef] [PubMed]
  17. Hara, A.; Radin, N.S. Lipid extraction of tissues with a low-toxicity solvent. Anal. Biochem. 1978, 90, 420–426. [Google Scholar] [CrossRef]
  18. Bainbridge, M.L.; Lock, A.L.; Kraft, J. Lipid-encapsulated echium oil (Echium plantagineum) increases the content of stearidonic acid in plasma lipid fractions and milk fat of dairy cows. J. Agric. Food Chem. 2015, 63, 4827–4835. [Google Scholar] [CrossRef]
  19. Unger, A.L.; Bourne, D.E.; Walsh, H.; Kraft, J. Fatty acid content of retail cow’s milk in the Northeastern United States-What’s in it for the consumer? J. Agric. Food Chem. 2020, 68, 4268–4276. [Google Scholar] [CrossRef]
  20. Wolff, R.L.; Bayard, C.C.; Fabien, R.J. Evaluation of sequential methods for the determination of butterfat fatty acid composition with emphasis ontrans-18:1 acids. Application to the study of seasonal variations in french butters. J. Am. Oil Chem. Soc. 1995, 72, 1471–1483. [Google Scholar] [CrossRef]
  21. Bainbridge, M.L.; Cersosimo, L.M.; Wright, A.-D.G.; Kraft, J. Rumen bacterial communities shift across a lactation in Holstein, Jersey and Holstein × Jersey dairy cows and correlate to rumen function, bacterial fatty acid composition and production parameters. FEMS Microbiol. Ecol. 2016, 92, fiw059. [Google Scholar] [CrossRef]
  22. Shingfield, K.J.; Reynolds, C.K.; Hervás, G.; Griinari, J.M.; Grandison, A.S.; Beever, D.E. Examination of the Persistency of Milk Fatty Acid Composition Responses to Fish Oil and Sunflower Oil in the Diet of Dairy Cows. J. Dairy Sci. 2006, 89, 714–732. [Google Scholar] [CrossRef] [PubMed]
  23. Herawati, N.; Sutrisno, A.; Nusyirwan, N.; Misgiyati, M. The performance of ridge regression, LASSO, and elastic-net in controlling multicollinearity: A simulation and application. J. Mod. Appl. Stat. Methods 2024, 23, 1–13. [Google Scholar] [CrossRef]
  24. Vlaeminck, B.; Fievez, V.; Cabrita, A.R.J.; Fonseca, A.J.M.; Dewhurst, R.J. Factors affecting odd- and branched-chain fatty acids in milk: A review. Anim. Feed Sci. Technol. 2006, 131, 389–417. [Google Scholar] [CrossRef]
  25. Kupczynski, R.; Pacyga, K.; Lewandowska, K.; Bednarski, M.; Szumny, A. Milk odd- and branched-chain fatty acids as biomarkers of rumen fermentation. Animals 2024, 14, 1706. [Google Scholar] [CrossRef]
  26. Rico, D.E.; Chouinard, P.Y.; Hassanat, F.; Benchaar, C.; Gervais, R. Prediction of enteric methane emissions from Holstein dairy cows fed various forage sources. Animal 2016, 10, 203–211. [Google Scholar] [CrossRef]
  27. van Lingen, H.J.; Crompton, L.A.; Hendriks, W.H.; Reynolds, C.K.; Dijkstra, J. Meta-analysis of relationships between enteric methane yield and milk fatty acid profile in dairy cattle. J. Dairy Sci. 2014, 97, 7115–7132. [Google Scholar] [CrossRef]
  28. Bayat, A.R.; Kairenius, P.; Stefanski, T.; Leskinen, H.; Comtet-Marre, S.; Forano, E.; Chaucheyras-Durand, F.; Shingfield, K.J. Effect of camelina oil or live yeasts (Saccharomyces cerevisiae) on ruminal methane production, rumen fermentation, and milk fatty acid composition in lactating cows fed grass silage diets. J. Dairy Sci. 2015, 98, 3166–3181. [Google Scholar] [CrossRef]
  29. Kramer, J.K.G.; Fellner, V.; Dugan, M.E.R.; Sauer, F.D.; Mossoba, M.M.; Yurawecz, M.P. Evaluating acid and base catalysts in the methylation of milk and rumen fatty acids with special emphasis on conjugated dienes and total trans fatty acids. Lipids 1997, 32, 1219–1228. [Google Scholar] [CrossRef]
  30. Marseglia, A.; Caligiani, A.; Comino, L.; Righi, F.; Quarantelli, A.; Palla, G. Cyclopropyl and omega-cyclohexyl fatty acids as quality markers of cow milk and cheese. Food Chem. 2013, 140, 711–716. [Google Scholar] [CrossRef] [PubMed]
  31. Kawashima, H.; Yoshizawa, K. The physiological and pathological properties of Mead acid, an endogenous multifunctional n-9 polyunsaturated fatty acid. Lipids Health Dis. 2023, 22, 172. [Google Scholar] [CrossRef] [PubMed]
  32. Cleland, K.A.; James, M.J.; Neumann, M.A.; Gibson, R.A.; Cleland, L.G. Differences in fatty acid composition of immature and mature articular cartilage in humans and sheep. Lipids 1995, 30, 949–953. [Google Scholar] [CrossRef] [PubMed]
  33. Ali, O.; Szabó, A. Review of eukaryote cellular membrane lipid composition, with special attention to the fatty acids. Int. J. Mol. Sci. 2023, 24, 15693. [Google Scholar] [CrossRef] [PubMed]
  34. Fiore, E.; Lisuzzo, A.; Tessari, R.; Spissu, N.; Moscati, L.; Morgante, M.; Gianesella, M.; Badon, T.; Mazzotta, E.; Berlanda, M.; et al. Milk fatty acids composition changes according to β-hydroxybutyrate concentrations in ewes during early lactation. Animals 2021, 11, 1371. [Google Scholar] [CrossRef] [PubMed]
  35. Garnsworthy, P.C.; Craigon, J.; Hernandez-Medrano, J.H.; Saunders, N. Variation among individual dairy cows in methane measurements made on farm during milking. J. Dairy Sci. 2012, 95, 3181–3189. [Google Scholar] [CrossRef]
  36. de Souza, J.; Leskinen, H.; Lock, A.L.; Shingfield, K.J.; Huhtanen, P. Between-cow variation in milk fatty acids associated with methane production. PLoS ONE 2020, 15, e0235357. [Google Scholar] [CrossRef]
Lipidology 02 00024 g001
Figure 1. Heatmap summary of Spearman’s correlation coefficients between milk fatty acids (% of total fatty acid methyl esters identified) and methane output (O; g/d), yield (Y; g/kg dry matter intake), and intensity (I; g/kg energy-corrected milk). Each square represents the result of a pair of variables. The red values indicate a positive relationship between two variables, while the blue values indicate an inverse relationship between variables. Beige colors indicate weak or no linear correlation. * Denotes p-values < 0.05. ** Denotes p-values < 0.01. *** Denotes p-values < 0.001. BCFAs, branched-chain fatty acids; MUFAs, monounsaturated fatty acids; OBCFAs, odd- and branched-chain fatty acids; OCFAs, odd-chain fatty acids; PUFAs, polyunsaturated fatty acids; SFAs, saturated fatty acids; SMCFAs, short- and medium-chain fatty acids; VLCFAs, very-long-chain fatty acids.
Figure 1. Heatmap summary of Spearman’s correlation coefficients between milk fatty acids (% of total fatty acid methyl esters identified) and methane output (O; g/d), yield (Y; g/kg dry matter intake), and intensity (I; g/kg energy-corrected milk). Each square represents the result of a pair of variables. The red values indicate a positive relationship between two variables, while the blue values indicate an inverse relationship between variables. Beige colors indicate weak or no linear correlation. * Denotes p-values < 0.05. ** Denotes p-values < 0.01. *** Denotes p-values < 0.001. BCFAs, branched-chain fatty acids; MUFAs, monounsaturated fatty acids; OBCFAs, odd- and branched-chain fatty acids; OCFAs, odd-chain fatty acids; PUFAs, polyunsaturated fatty acids; SFAs, saturated fatty acids; SMCFAs, short- and medium-chain fatty acids; VLCFAs, very-long-chain fatty acids.
Lipidology 02 00024 g001
Table 1. Ingredient composition for the high-yield, low-yield, and primiparous total mixed rations (TMRs) on a dry matter (DM) basis.
Table 1. Ingredient composition for the high-yield, low-yield, and primiparous total mixed rations (TMRs) on a dry matter (DM) basis.
Ingredient, % DMHigh-Yield TMRLow-Yield TMRPrimiparous TMR
Corn silage49.4434.5147.74
Haylage10.8527.809.91
High-grain mix39.70-42.34
  Fine corn meal7.23-7.72
  Canola meal4.82-5.14
  Gluten feed2.25-2.40
  Soybean meal6.11-6.51
  Calcium salt fat0.82-0.88
  Cane molasses1.10-1.17
  Calcium carbonate1.25-1.33
  C16 fat1.27-1.35
  SQ_810 10.75-0.80
  Salt0.37-0.39
  Magnesium oxide0.29-0.31
  Mineral/vitamin premix0.11-0.11
  XPC yeast culture 20.05-0.05
  Smartamine 30.05-0.05
  Urea0.40-0.43
  Bakery feed3.54-3.77
  Intergral A+ 40.05-0.05
  Steam flaked corn1.61-1.71
  Distiller grains1.21-1.29
  Dry MHA 50.09-0.10
  PGI amino enhancer0.73-0.78
  Amino max PGI5.63-6.00
Low-grain mix-37.69-
  Fine corn meal-15.11-
  Soybean meal-5.88-
  Canola meal-6.71-
  Calcium carbonate-1.25-
  Cane molasses-0.87-
  C16 fat-0.88-
  SQ_810 1-0.45-
  Salt-0.41-
  Magnesium oxide-0.28-
  Mineral/vitamin premix-0.11-
  Liquid Metasmart 3-0.07-
  XPC yeast culture 2-0.05-
  Amino max PGI-5.46-
  Urea-0.17-
SQ_810, sodium sesquicarbonate. 1 Arm & Hammer Animal Nutrition (Ewing, NJ, USA). 2 Diamond V (Cedar Rapids, IA, USA). 3 Adisseo (Alpharetta, GA, USA). 4 Alltech (Essex Junction, VT, USA). 5 Novus (Chesterfield, MO, USA).
Table 2. Nutrient contents of the high-yield, low-yield, and first-calf heifer total mixed rations (TMRs).
Table 2. Nutrient contents of the high-yield, low-yield, and first-calf heifer total mixed rations (TMRs).
ItemHigh-Yield TMRLow-Yield TMRPrimiparous TMR
DM, %39.1541.1543.05
CP, % DM15.7816.1016.45
Soluble protein, % CP39.2543.0036.50
ADF, % DM21.0825.6020.60
aNDFom, % DM33.4036.4532.60
Lignin, % DM3.404.503.50
Crude fiber, % DM13.0015.4011.25
NFC, % DM39.4535.6038.75
Starch, % DM25.7022.5526.65
Ether extract, % DM5.664.905.62
Simple sugar, % DM4.484.804.80
Calcium, % DM0.660.760.83
Phosphorus, % DM0.380.350.40
Magnesium, % DM0.330.320.35
Potassium, % DM1.001.161.02
Sodium, % DM0.350.250.36
Iron, mg/kg271267252
Zinc, mg/kg51.544.554.0
Copper, mg/kg12.011.012.5
Manganese, mg/kg47.042.041.5
Molybdenum, mg/kg1.101.251.30
NEL, Mcal/kg1.781.651.76
NEM, Mcal/kg1.821.651.80
NEG, Mcal/kg1.191.041.18
ADF, acid detergent fiber; aNDFom, ash-free neutral detergent fiber; CP, crude protein; DM, dry matter; NEG, net energy for growth; NEL, net energy for lactation; NEM, net energy for maintenance; NFC, non-fiber carbohydrate.
Table 3. Descriptive statistics outlining the production performance and methane emission parameters of lactating Holstein cows averaged over the observation period (n = 40).
Table 3. Descriptive statistics outlining the production performance and methane emission parameters of lactating Holstein cows averaged over the observation period (n = 40).
VariableMeanMedianMinimumMaximumSDSEM
Body weight, kg 179380567095775.4311.93
Body condition score 13.283.252.503.880.270.04
Dry matter intake, kg/d25.826.320.330.42.200.35
Neutral detergent fiber, kg/d (% of DM)8.348.616.269.820.830.13
Starch, kg/d (% of DM)6.316.285.147.900.550.09
Ether extract, kg/d (% of DM)1.371.361.091.630.130.02
Days in milk1901862831791.214.42
Milk yield, kg/d40.638.727.160.18.351.32
Energy-corrected milk 2, kg/d41.338.428.162.38.501.34
Feed efficiency 31.611.561.052.630.350.06
Fat, %4.254.342.925.470.530.08
Fat yield, kg/d1.721.601.072.750.410.06
Protein, %3.253.262.643.850.280.04
Protein yield, kg/d1.311.250.891.780.220.04
CH4 output, g/d366.1357.8238.2515.972.2911.43
CH4 yield, g/kg DMI14.113.710.119.22.260.36
CH4 intensity, g/kg ECM9.409.443.8318.333.280.52
1 Measurements taken at the beginning of the trial period. 2 Calculated as per Engelke et al. [7]: ECM (kg/d) = [1.05 + 0.38 × milk fat (%) + 0.21 × milk protein (%)]/3.28 × milk yield (kg/d). 3 Calculated as energy-corrected milk/dry matter intake. CH4, methane; DMI, dry matter intake; ECM, energy-corrected milk.
Table 4. Comprehensive descriptive statistics for milk fatty acid composition (% of total fatty acid methyl esters) of lactating Holstein cows (n = 40).
Table 4. Comprehensive descriptive statistics for milk fatty acid composition (% of total fatty acid methyl esters) of lactating Holstein cows (n = 40).
Fatty AcidMeanMedianMinimumMaximumSDSEM
4:03.344.122.703.280.310.05
5:00.040.070.020.040.010.00
6:02.092.511.562.060.210.03
7:00.030.070.020.030.010.00
8:01.111.350.811.100.140.02
9:00.040.080.020.030.010.00
10:02.342.961.442.370.370.06
11:00.120.170.060.110.030.00
11-cyclohexyl-11:00.140.220.080.130.030.01
12:02.753.411.582.800.460.07
13:0-iso0.020.040.010.020.010.00
13:0-anteiso0.070.090.030.070.020.00
13:00.170.260.100.170.040.01
14:0-iso0.070.130.040.070.020.00
14:09.9611.956.6210.311.200.19
14:1 t90.010.020.000.010.000.00
14:1 c90.891.300.480.890.200.03
15:0-iso0.190.270.130.190.030.01
15:0-anteiso0.380.500.260.380.060.01
15:01.001.370.650.970.170.03
16:0-iso0.180.300.110.170.060.01
16:036.0742.1830.2835.932.430.38
16:1 t90.040.050.010.040.010.00
16:1 c70.030.060.020.030.010.00
16:1 c80.180.240.130.170.030.00
16:1 c91.682.541.101.610.310.05
16:1 c10/t130.010.030.000.010.000.00
17:0-iso0.280.340.210.290.030.00
17:0-anteiso0.360.460.280.360.040.01
17:00.530.610.420.530.050.01
17:1 c70.030.040.010.030.010.00
17:1 c90.180.380.110.160.060.01
18:0-iso/17:1 t100.040.080.010.040.010.00
18:08.9511.996.588.821.270.20
18:1 t40.020.030.010.020.010.00
18:1 t50.020.040.010.020.010.00
18:1 t6-t80.260.360.170.270.050.01
18:1 t90.240.310.170.240.030.01
18:1 t100.410.810.230.410.120.02
18:1 t110.741.190.470.730.130.02
18:1 t120.320.450.220.310.060.01
18:1 t150.160.230.100.160.030.00
18:1 c6-c8/t13/t140.730.900.460.740.110.02
18:1 c918.7327.5214.6917.943.060.48
18:1 c110.661.040.430.610.150.02
18:1 c120.290.440.180.270.060.01
18:1 c130.070.140.040.060.030.00
18:1 c14/t160.240.310.190.240.030.00
18:1 c15/19:00.130.150.100.130.010.00
18:1 c160.060.080.030.060.010.00
18:2 t10,t140.030.050.000.030.010.00
18:2 c5,t13/t8,ct120.190.290.120.180.040.01
18:2 c9,t110.380.530.250.380.070.01
18:2 c9,t140.090.130.050.090.020.00
18:2 c12,t160.040.060.010.040.010.00
18:2 c9,c121.872.381.261.890.240.04
18:3 c6,c9,c120.030.050.020.030.010.00
18:3 c9,c12,c150.260.400.180.250.060.01
20:00.110.140.090.110.010.00
20:1 c90.090.120.070.090.010.00
20:1 11c0.040.060.020.030.010.00
20:2 c11,c140.020.040.010.020.010.00
20:3 c5,c8,c110.120.190.050.110.030.01
20:4 c5,c8,c11,c140.120.170.080.120.020.00
20:5 c5,c8,c11,c14,c170.030.050.020.030.010.00
22:00.030.070.020.030.010.00
22:4 c7,c10,c13,c160.030.060.000.020.010.00
22:5 c7,c10,c13,c16,c190.050.080.030.050.010.00
23:00.020.040.000.020.010.00
24:00.020.040.000.020.010.00
∑ SFAs69.9676.1160.2870.673.550.56
∑ SMCFAs 121.5825.7915.3521.832.260.36
∑ OBCFAs3.904.593.233.880.360.06
∑ OCFAs 22.152.751.692.100.260.04
∑ BCFAs1.752.301.351.680.240.04
∑ iso-BCFAs0.801.070.640.770.120.02
∑ anteiso-BCFAs0.891.140.680.870.110.02
∑ MUFAs26.2535.7420.9225.313.410.54
∑ trans-MUFAs20.9029.7716.5019.993.120.49
∑ cis-MUFAs20.9029.8716.5719.953.170.50
∑ 18:1 isomers23.0632.1218.0922.053.250.51
∑ PUFAs3.254.062.393.280.340.05
∑ 18:2 isomers2.593.401.742.620.330.05
∑ 18:3 isomers0.290.430.200.280.060.01
∑ VLCFAs 30.670.890.470.670.090.01
1 Includes 4:0, 6:0, 8:0, 10:0, and 12:0. 2 Includes 5:0, 7:0, 9:0, 11:0, 13:0, 15:0, and 17:0. 3 Fatty acids with a chain length ≥ 20 carbons. BCFAs, branched-chain fatty acids, MUFAs, monounsaturated fatty acids; OBCFAs, odd- and branched-chain fatty acids; OCFAs, odd-chain fatty acids; PUFAs, polyunsaturated fatty acids; SFAs, saturated fatty acids; SMCFAs, short- and medium-chain fatty acids.
Table 5. Comprehensive Spearman correlation coefficients between methane emission metrics and milk fatty acids (% of total fatty acid methyl esters identified) and production parameters. Values were calculated on individual values (n = 40).
Table 5. Comprehensive Spearman correlation coefficients between methane emission metrics and milk fatty acids (% of total fatty acid methyl esters identified) and production parameters. Values were calculated on individual values (n = 40).
VariableCH4 Output,
g/d		CH4 Yield,
g/kg DMI		CH4 Intensity,
g/kg ECM
rp-Valuerp-Valuerp-Value
Days in milk0.62<0.0010.64<0.0010.70<0.001
Dry matter intake, kg/d0.62<0.0010.260.110.330.04
Energy-corrected milk 1, kg/d−0.46<0.01−0.64<0.001−0.81<0.001
aNDFom, kg/d (% of DM)0.67<0.0010.320.040.420.01
Starch, kg/d (% of DM)0.420.010.080.640.140.38
Milk yield, kg/d−0.55<0.001−0.72<0.001−0.84<0.001
Feed efficiency 2−0.72<0.001−0.72<0.001−0.93<0.001
4:0−0.020.93−0.110.50−0.060.69
5:0−0.240.13−0.360.02−0.420.01
6:00.400.010.270.090.230.15
7:0−0.010.98−0.150.36−0.190.23
8:00.48<0.010.370.020.320.04
9:00.050.74−0.080.64−0.150.35
10:00.48<0.010.360.020.310.05
11:00.350.030.330.040.300.06
11-cyclohexyl-11:00.72<0.0010.61<0.0010.68<0.001
12:00.48<0.010.400.010.350.03
13:0-iso0.66<0.0010.59<0.0010.65<0.001
13:0-anteiso0.53<0.0010.59<0.0010.51<0.01
13:00.340.030.250.120.190.25
14:0-iso0.73<0.0010.64<0.0010.75<0.001
14:00.51<0.010.400.010.46<0.01
14:1 t90.41<0.010.410.010.380.02
14:1 c90.290.070.470.000.410.01
15:0-iso0.66<0.0010.62<0.0010.73<0.001
15:0-anteiso0.55<0.0010.58<0.0010.63<0.001
15:00.320.050.210.190.160.32
16:0-iso0.69<0.0010.63<0.0010.71<0.001
16:00.290.060.270.090.190.25
16:1 t9−0.350.03−0.260.11−0.240.14
16:1 c7−0.180.26−0.230.14−0.180.26
16:1 c8−0.62<0.001−0.56<0.001−0.50<0.01
16:1 c9−0.340.03−0.210.20−0.290.07
16:1 c10/t130.230.150.030.880.030.85
17:0-iso0.080.620.150.350.220.17
17:0-anteiso0.200.210.230.150.300.06
17:00.070.660.080.620.040.82
17:1 c70.420.010.45<0.010.430.01
17:1 c9−0.54<0.001−0.410.01−0.47<0.01
18:0-iso/17:1 t10−0.320.04−0.200.22−0.150.36
18:0−0.250.12−0.230.16−0.210.20
18:1 t4−0.040.83−0.170.28−0.110.51
18:1 t5−0.110.50−0.240.13−0.270.09
18:1 t6-t8−0.220.16−0.290.07−0.290.07
18:1 t9−0.180.28−0.280.08−0.270.09
18:1 t10−0.380.02−0.440.00−0.44<0.01
18:1 t11−0.140.37−0.210.19−0.200.22
18:1 t12−0.110.50−0.220.18−0.200.21
18:1 t150.090.59−0.050.76−0.070.67
18:1 c6-c8/t13/t140.160.320.020.880.040.79
18:1 c9−0.420.01−0.290.07−0.240.14
18:1 c11−0.77<0.001−0.66<0.001−0.67<0.001
18:1 c120.050.76−0.130.42−0.130.42
18:1 c13−0.54<0.001−0.460.00−0.400.01
18:1 c14/t160.050.77−0.060.690.060.70
18:1 c15/19:0−0.030.86−0.110.51−0.010.96
18:1 c16−0.250.12−0.260.11−0.220.17
18:2 t10,t140.280.080.270.100.300.06
18:2 c5,t13/t8,ct12−0.050.77−0.080.62−0.030.85
18:2 c9,t11−0.250.12−0.190.25−0.170.29
18:2 c9,t14−0.060.73−0.030.860.010.94
18:2 c12,t16−0.120.45−0.190.24−0.140.38
18:2 c9,c120.050.780.060.720.170.30
18:3 c6,c9,c120.050.75−0.040.790.040.79
18:3 c9,c12,c150.280.080.260.100.380.02
20:00.260.100.190.240.280.08
20:1 c90.410.010.48<0.010.56<0.001
20:1 11c−0.410.01−0.370.02−0.290.07
20:2 c11,c140.45<0.010.420.010.62<0.001
20:3 c5,c8,c110.68<0.0010.58<0.0010.65<0.001
20:4 c5,c8,c11,c140.53<0.0010.520.000.55<0.001
20:5 c5,c8,c11,c14,c170.190.240.110.510.200.23
22:00.52<0.010.47<0.010.58<0.001
22:4 c7,c10,c13,c160.59<0.0010.55<0.0010.64<0.001
22:5 c7,c10,c13,c16,c190.310.050.370.020.48<0.01
23:00.46<0.010.45<0.010.53<0.001
24:00.60<0.0010.51<0.010.62<0.001
∑ SFAs0.47<0.010.350.030.310.05
∑ SMCFAs 30.58<0.0010.44<0.010.45<0.01
∑ OBCFAs0.56<0.0010.50<0.010.47<0.01
∑ OCFAs 40.190.230.110.500.060.72
∑ BCFAs0.63<0.0010.62<0.0010.72<0.001
∑ iso-BCFAs0.62<0.0010.60<0.0010.72<0.001
∑ anteiso-BCFAs0.56<0.0010.57<0.0010.64<0.001
∑ MUFAs−0.51<0.01−0.360.02−0.330.04
∑ 18:1 isomers−0.48<0.01−0.380.02−0.330.04
∑ trans-isomers−0.47<0.01−0.360.02−0.300.06
∑ cis-isomers−0.46<0.01−0.350.03−0.300.06
∑ PUFAs0.070.660.020.900.100.52
∑ 18:2 isomers−0.150.37−0.190.25−0.130.42
∑ 18:3 isomers0.270.090.230.160.360.02
∑ VLCFAs 50.69<0.0010.65<0.0010.79<0.001
1 Calculated as per Engelke et al. [7]: ECM (kg/d) = [1.05 + 0.38 × milk fat (%) + 0.21 × milk protein (%)]/3.28 × milk yield (kg/d). 2 Calculated as energy-corrected milk/dry matter intake. 3 Includes 4:0, 6:0, 8:0, 10:0, and 12:0. 4 Includes 5:0, 7:0, 9:0, 11:0, 13:0, 15:0, and 17:0. 5 Fatty acids with a chain length ≥ 20 carbons. aNDFom, ash-free neutral detergent fiber; BCFAs, branched-chain fatty acids, CH4, methane; MUFAs, monounsaturated fatty acids; OBCFAs, odd- and branched-chain fatty acids; OCFAs, odd-chain fatty acids; PUFAs, polyunsaturated fatty acids; SFAs, saturated fatty acids; SMCFAs, short- and medium-chain fatty acids; VLCFAs, very-long-chain fatty acids.
Table 6. Performance summary for methane output, yield, and intensity across three different regression models: ridge, LASSO, and elastic net.
Table 6. Performance summary for methane output, yield, and intensity across three different regression models: ridge, LASSO, and elastic net.
Response VariableModelRegression TypeLambdaR2CCCMSERMSERMSPE
CH4 output,
g/d		1Ridge4.770.830.90861.129.30.08
LASSO0.060.850.92755.727.50.08
Elastic net0.430.850.92760.427.60.08
2Ridge21.80.810.89952.930.90.09
LASSO0.320.830.90888.029.80.08
Elastic net4.160.820.90895.329.90.08
3Ridge9.440.810.89991.831.50.09
LASSO0.100.820.90933.630.60.09
Elastic net7.270.790.871067.832.70.09
CH4 yield,
g/kg DMI		1Ridge0.160.760.861.201.100.08
LASSO0.020.780.871.121.060.08
Elastic net0.000.780.871.111.050.08
2Ridge0.450.710.821.441.200.09
Lasso0.010.720.831.411.190.09
Elastic net0.120.720.831.421.190.09
3Ridge0.960.710.821.441.200.09
LASSO0.100.720.831.391.180.09
Elastic net0.260.720.821.411.190.09
CH4 intensity,
g/kg ECM		1Ridge0.290.870.931.381.180.14
LASSO0.030.870.931.321.150.15
Elastic net0.170.870.931.411.190.14
2Ridge0.640.800.902.081.400.15
LASSO0.020.820.911.911.380.14
Elastic net0.020.820.911.911.380.14
3Ridge0.710.800.882.101.450.16
LASSO0.050.820.901.901.380.15
Elastic net0.020.820.901.871.370.15
CCC, concordance correlation coefficient; CH4, methane; DMI, dry matter intake; ECM, energy-corrected milk; MSE, mean square error; RMSE, root mean square error; RMSPE, root mean square prediction error.
Table 7. Standardized coefficients and relative variance of input variables for methane (CH4) output, yield, and intensity regression models.
Table 7. Standardized coefficients and relative variance of input variables for methane (CH4) output, yield, and intensity regression models.
Response VariableModelPredictorRidge
Coefficient (±SD) 1		RVLASSO Coefficient (±SD) 1RVElastic Net Coefficient (±SD) 1RVDirection
CH4 output, g/d1FA and intake variables 2
 13:0-anteiso1245.4 ± 344.90.281663.9 ± 463.60.281579.8 ± 452.70.29+
 14:0−5.8 ± 4.90.84−12.8 ± 6.70.52−11.4 ± 6.80.59-
 ∑ iso-BCFAs267.7 ± 57.80.22395.6 ± 95.70.24374.1 ± 80.50.22+
 Milk yield−2.0 ± 0.70.37−1.2 ± 0.90.81−1.3 ± 0.90.70-
 NDF intake47.6 ± 8.10.1758.8 ± 11.10.1956.9 ± 11.00.19+
Categorical variables 3
 High-yield diet (vs. Primiparous)−16.8 ± 13.00.77−39.9 ± 18.80.47−35.4 ± 19.80.56-
 Low-yield diet (vs. Primiparous)−51.2 ± 17.10.33−97.0 ± 26.40.27−89.0 ± 26.20.29-
2FA and intake variables 2
 13:0-anteiso666.5 ± 305.60.46665.3 ± 396.90.60646.6 ± 361.90.56+
 16:0-iso231.0 ± 88.70.38304.6 ± 263.80.87274.9 ± 194.80.71+
 11-cyclohexyl-11:0467.6 ± 189.50.41484.7 ± 4080.84484.6 ± 336.40.69+
 ∑ VLCFAs147.9 ± 55.20.37130.4 ± 95.70.73133.9 ± 85.80.64+
 Milk yield−1.5 ± 0.60.44−1.6 ± 1.30.80−1.6 ± 1.00.63-
 NDF intake31.0 ± 7.70.2541.6 ± 12.60.3039.0 ± 11.90.31+
Categorical variables 3
 High-yield diet (vs. Primiparous)−3.2 ± 10.53.30−18.7 ± 21.51.15−12.8 ± 19.21.50-
 Low-yield diet (vs. Primiparous)−25.6 ± 12.90.51−53.9 ± 25.10.47−44.7 ± 22.30.50-
3FA and intake variables 2
 11-cyclohexyl-11:0759.2 ± 202.50.27866.0 ± 307.70.36682.2 ± 220.60.32+
 20:4 c5,c8,c11,c14679.4 ± 309.30.46605.6 ± 410.80.68621.4 ± 349.80.56+
 Milk yield−2.3 ± 0.70.32−2.5 ± 1.20.47−2.1 ± 0.80.38-
 NDF intake37.3 ± 7.30.2046.9 ± 11.00.2334.4 ± 9.60.28+
Categorical variables 3
 High-yield diet (vs. Primiparous)−5.6 ± 10.81.93−18.8 ± 21.71.16DroppedN/A-
 Low-yield diet (vs. Primiparous)−13.6 ± 16.11.18−37.2 ± 28.30.76DroppedN/A-
CH4 yield,
g/kg DMI		1FA and intake variables 2
 13:0-anteiso41.05 ± 11.10.2747.2 ± 14.30.3050.2 ± 14.00.28+
 15:0-iso29.9 ± 8.80.2938.3 ± 11.80.3141.2 ± 11.40.28+
 16:0-iso17.6 ± 5.80.3321.2 ± 7.60.3622.5 ± 6.80.30+
 20:3 c5,c8,c1114.6 ± 6.10.4413.1 ± 7.20.5512.9 ± 7.80.60+
 ∑ SMCFAs−0.3 ± 0.10.38−0.3 ± 0.10.32−0.4 ± 0.10.30-
Categorical variables 3
 High-yield diet (vs. Primiparous)−0.4 ± 0.51.18−0.4 ± 0.51.41−0.4 ± 0.71.61-
 Low-yield diet (vs. Primiparous)−1.5 ± 0.80.51−2.2 ± 0.90.42−2.6 ± 0.90.35-
2FA and intake variables 2
 13:0-anteiso21.0 ± 9.80.4718.5 ± 14.20.7717.7 ± 11.60.66+
 11-cyclohexyl-11:024.4 ± 5.30.2228.2 ± 6.70.2426.9 ± 6.70.25+
 20:4 c5,c8,c11,c1421.4 ± 9.70.4522.5 ± 13.10.5820.8 ± 12.60.61+
 Milk yield−0.1 ± 0.020.24−0.1 ± 0.030.36−0.1 ± 0.030.32+
3FA and intake variables 2
 15:0-iso9.8 ± 4.10.425.3 ± 9.11.737.4 ± 7.51.02+
 18:1 c11−2.0 ± 0.90.43−1.3 ± 1.51.23−1.3 ± 1.41.05-
 11-cyclohexyl-11:015.5 ± 4.10.2622.5 ± 9.70.4319.7 ± 7.50.38+
 20:4 c5,c8,c11,c1414.8 ± 7.50.5116.6 ± 11.90.7114.4 ± 11.80.82+
 Milk yield−0.1 ± 0.020.25−0.1 ± 0.030.45−0.1 ± 0.030.41-
Categorical variables 3
 Mid-Lactation (vs. Early)0.6 ± 0.30.460.5 ± 0.50.860.5 ± 0.40.81+
 Late Lactation (vs. Early)−0.1 ± 0.22.26DroppedN/ADroppedN/A-
CH4
intensity,
g/kg ECM		1FA and intake variables 2
 15:0-iso19.2 ± 8.20.4324.1 ± 14.60.6016.3 ± 10.10.62+
 16:0-iso15.8 ± 4.80.3016.5 ± 6.30.3813.1 ± 5.40.41+
 ∑ VLCFAs3.4 ± 2.70.81DroppedN/A1.8 ± 2.81.57+
Categorical variables 3
 Feed efficiency−4.5 ± 0.90.19−5.5 ± 1.20.21−5.2 ± 1.00.20-
 High-yield diet (vs. Primiparous)−0.2 ± 0.41.83−0.03 ± 0.413.03DroppedN/A-
 Low-yield diet (vs. Primiparous)−0.6 ± 0.61.10−0.8 ± 0.91.09DroppedN/A-
2FA and intake variables 2
 16:0-iso26.6 ± 6.10.2329.0 ± 6.70.2329.0 ± 6.50.22+
 ∑ SMCFAs−0.2 ± 0.10.55−0.4 ± 0.10.38−0.4 ± 0.20.43-
 ∑ VLCFAs15.5 ± 3.00.2015.7 ± 3.90.2515.7 ± 3.80.24+
Categorical variables 3
 Mid-lactation (vs. Early)2.0 ± 0.540.272.6 ± 0.80.322.6 ± 0.80.29+
 Late lactation (vs. Early)1.7 ± 0.50.302.3 ± 0.70.302.3 ± 0.70.31+
3FA and intake variables 2
 15:0-iso25.6 ± 8.00.3135.6 ± 17.30.4935.8 ± 17.50.49+
 16:0-iso16.3 ± 3.90.2421.6 ± 9.00.4221.4 ± 7.90.37+
 11-cyclohexyl-11:013.7 ± 7.00.511.8 ± 14.58.022.3 ± 12.25.32+
 20:1 c935.0 ± 23.60.6725.5 ± 36.31.4325.5 ± 39.71.56+
 20:3 c5,c8,c1125.4 ± 5.70.2333.7 ± 9.00.2733.8 ± 8.50.25+
 ∑ SMCFAs−0.2 ± 0.10.47−0.4 ± 0.20.40−0.4 ± 0.20.4-
 Days in milk0.004 ± 0.0030.750.01 ± 0.011.000.01 ± 0.011.00+
1 Results derived from 200 bootstrap replicates using 10-fold cross-validation. 2 Reported values: Fatty acids, % of total fatty acid methyl esters identified; milk yield, kg/d; N/A, not applicable; NDF intake, kg/d (% of dry matter). 3 Baseline levels are the primiparous diet and early lactation stage. CH4, methane; DMI, dry matter intake; RV, relative variance; SD, standard deviation; SMCFAs, short- and medium-chain fatty acids; VLCFAs, very-long-chain fatty acids; +, positive correlation; -, negative correlation.
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Share and Cite

MDPI and ACS Style

Youngmark, E.C.; Kraft, J. Milk Fatty Acid Profiling as a Tool for Estimating Methane Emissions in Conventionally Fed Dairy Cows. Lipidology 2025, 2, 24. https://doi.org/10.3390/lipidology2040024

AMA Style

Youngmark EC, Kraft J. Milk Fatty Acid Profiling as a Tool for Estimating Methane Emissions in Conventionally Fed Dairy Cows. Lipidology. 2025; 2(4):24. https://doi.org/10.3390/lipidology2040024

Chicago/Turabian Style

Youngmark, Emily C., and Jana Kraft. 2025. "Milk Fatty Acid Profiling as a Tool for Estimating Methane Emissions in Conventionally Fed Dairy Cows" Lipidology 2, no. 4: 24. https://doi.org/10.3390/lipidology2040024

APA Style

Youngmark, E. C., & Kraft, J. (2025). Milk Fatty Acid Profiling as a Tool for Estimating Methane Emissions in Conventionally Fed Dairy Cows. Lipidology, 2(4), 24. https://doi.org/10.3390/lipidology2040024

Article Metrics

Back to TopTop