Discovery of different metabotypes in overconditioned dairy cows by means of machine learning

Liver transcriptome profiles of dairy cows with different serum metabotypes

In a previously established animal model, 38 multiparous Holstein cows were assigned to 2 groups fed different diets to achieve either a normal (NBCS) or high (HBCS) body condition score (BCS) and backfat thickness (BFT) until dry-off at -49 d before calving (NBCS: BCS <3.5 [3.02 ± 0.24) and BFT <1.2 cm [0.92 ± 0.21]; HBCS: BCS >3.75 [3.82 ± 0.33] and BFT >1.4 cm [2.36 ± 0.35], mean ± SD). The groups were also stratified for comparable milk yields (NBCS: 10,361 ± 302 kg; HBCS: 10,315 ± 437 kg; mean ± SD). The cows were then fed the same diet during the dry period and subsequent lactation, maintaining the differences in BFT and BCS throughout the study. Using the serum metabolomics data, we created a classification model that identified different metabotypes. Machine learning classifiers revealed a distinct cluster labeled HBCS-PN (HBCS predicted normal BCS) among over-conditioned cows. These cows showed higher feed intake and better energy balance than the HBCS-PH (high BCS predicted high BCS) group, while milk yield was similar. The aim of this study was to investigate the changes in the hepatic transcriptome of cows differing in serum-metabotype postpartum. We performed hepatic transcriptome analysis in cows from 3 metabolic clusters: HBCS-PH (n = 8), HBCS-PN (n = 6), and normal BCS predicted normal BCS (NBCS-PN, n = 8) on d 21 (±2) postpartum. Liver tissue from cows expressed a total of 13,118 genes aligned with the bovine genome. A total of 48 differentially expressed genes (DEG; false discovery rate ≤0.1 and fold-change >1.5) were found between NBCS-PN and HBCS-PH cows, whereas 24 DEG (14 downregulated and 10 upregulated) were found between HBCS-PN and HBCS-PH cows. The downregulated DEG (n = 31) in NBCS-PN cows compared with HBCS-PH cows are involved in biosynthetic processes such as lipid, lipoprotein, and cholesterol synthesis (e.g., APOA1, MKX, RPL3L, CANT1, CHPF, FUT1, ZNF696), cell organization, biogenesis, and localization (e.g., SLC12A8, APOA1, BRME1, RPL3L, STAG3, FBXW5, TMEM120A, SLC16A5, FGF21), catabolic processes (e.g., BREH1, MIOX, APOBEC2, FBXW5, NUDT16), and response to external stimuli (e.g., APOA1, FGF21, TMEM120A, FNDC4), whereas upregulated DEG (n = 17) are related to signal transduction and cell motility (e.g., RASSF2, ASPN, SGK1, KIF7, ZEB2, MAOA, ACKR4, TCAF1), suggesting altered metabolic adaptations during lactation. Our results showed 24 DEG between HBCS-PN and HBCS-PH in the liver. The expression of SLC12A8, SLC16A5, FBXW5, OSGIN1, LAMA3, KDELR3, OR4X17, and INHBE, which are responsible for regulating cellular processes was downregulated in HBCS-PN cows compared with HBCS-PH cows. In particular, the downregulation of SLC12A8 and SLC16A5 expression in HBCS-PN cows indicates lower metabolic load and reduced need for NAD+ biosynthesis to support mitochondrial respiratory processes. The upregulation of MAOA, ACKR4, KIF27, SFRP1, and CAV2 in the liver of HBCS-PN cows may indicate adaptive mechanisms to maintain normal liver function in response to increased metabolic demands from over-conditioning. These molecular differences underscore the existence of distinct metabolic types in cows and provide evidence for the role of the liver in shaping different metabolic patterns.

Invited review: Assessment of body condition score and body fat reserves in relation to insulin sensitivity and metabolic phenotyping in dairy cows

The purpose of this article is to review body condition scoring and the role of body fat reserves in relation to insulin sensitivity and metabolic phenotyping. This article summarizes body condition scoring assessment methods and the differences between subcutaneous and visceral fat depots in dairy cows. The mass of subcutaneous and visceral adipose tissue (AT) changes significantly during the transition period; however, metabolism and intensity of lipolysis differ between subcutaneous and visceral AT depots of dairy cows. The majority of studies on AT have focused on subcutaneous AT, and few have explored visceral AT using noninvasive methods. In this systematic review, we summarize the relationship between body fat reserves and insulin sensitivity and integrate omics research (e.g., metabolomics, proteomics, lipidomics) for metabolic phenotyping of cows, particularly overconditioned cows. Several studies have shown that AT insulin resistance develops during the prepartum period, especially in overconditioned cows. We discuss the role of AT lipolysis, fatty acid oxidation, mitochondrial function, acylcarnitines, and lipid insulin antagonists, including ceramide and glycerophospholipids, in cows with different body condition scoring. Nonoptimal body conditions (under- or overconditioned cows) exhibit marked abnormalities in metabolic and endocrine function. Overall, reducing the number of cows with nonoptimal body conditions in herds seems to be the most practical solution to improve profitability, and dairy farmers should adjust their management practices accordingly.

The Early Prediction of Common Disorders in Dairy Cows Monitored by Automatic Systems with Machine Learning Algorithms

We use multidimensional data from automated monitoring systems and milking systems to predict disorders of dairy cows by employing eight machine learning algorithms. The data included the season, days in milking, parity, age at the time of disorders, milk yield (kg/day), activity (unitless), six variables related to rumination time, and two variables related to the electrical conductivity of milk. We analyze 131 sick cows and 149 healthy cows with identical lactation days and parity; all data are collected on the same day, which corresponds to the diagnosis day for disordered cows. For disordered cows, each variable, except the ratio of rumination time from daytime to nighttime, displays a decreasing/increasing trend from d-7 or d-3 to d0 and/or d-1, with the d0, d-1, or d-2 values reaching the minimum or maximum. The test data sensitivity for three algorithms exceeded 80%, and the accuracies of the eight algorithms ranged from 65.08% to 84.21%. The area under the curve (AUC) of the three algorithms was >80%. Overall, Rpart best predicts the disorders with an accuracy, precision, and AUC of 81.58%, 92.86%, and 0.908, respectively. The machine learning algorithms may be an appropriate and powerful decision support and monitoring tool to detect herds with common health disorders.

Identification and characterization of dairy cows with different backfat thickness antepartum in relation to postpartum loss of backfat thickness: A cluster analytic approach

The objectives of this study were (1) to characterize the interindividual variation in the relationship between antepartum (ap) backfat thickness (BFT) and subsequent BFT loss during early lactation in a large dairy herd using cluster analysis; (2) to compare the serum concentrations of metabolites (nonesterified fatty acids, β-hydroxybutyrate), metabolic hormones (leptin and adiponectin), and an inflammatory marker (haptoglobin) among the respective clusters; and (3) to compare lactation performance and uterine health status in the different clusters. An additional objective was (4) to investigate differences in these serum variables and in milk yield of overconditioned (OC) cows that differed in the extent of BFT loss. Using data from a large study of 1,709 multiparous Holstein cows, we first selected those animals from which serum samples and BFT results (mm) were available at d 25 (±10) ap and d 31 (±3 d) postpartum (pp). The remaining 713 cows (parity of 2 to 7) were then subjected to cluster analysis: different approaches based on the BFT of the cows were performed. K-means (unsupervised machine learning algorithm) clustering based on BFT-ap alone identified 5 clusters: lean (5-8 mm BFT, n = 50), normal (9-12 mm, n = 206), slightly fat (SF; 13-16 mm, n = 203), just fat (JF; 16-22 mm, n = 193), and very fat (VF; 23-43 mm, n = 61). Clustering by difference between BFT-ap and BFT-pp (ΔBFT) also revealed 5 clusters: extreme loss (17-23 mm ΔBFT, n = 16), moderate loss (9-15 mm, n = 119), little loss (4-8 mm, n = 326), no loss (0-3 mm, n = 203), and gain (-8 to -1 mm, n = 51). Based on the blood variables measured, our results confirm that cows with greater BFT losses had higher lipid mobilization and ketogenesis than cows with less BFT loss. The serum variables of cows that gained BFT did not differ from normal cows. Milk yield was affected by the BFT-ap cluster, but not by the ΔBFT cluster. Cows categorized as VF had lesser milk yield than other clusters. We further compared the OC cows that had little or no BFT loss (i.e., 2% of VF, 12% of JF, and 31% of SF, OC-no loss, n = 85) with the OC cows that lost BFT (OC-loss, n = 135). Both NEFA and BHB pp concentrations and milk yield were greater in OC-loss cows compared with the OC-no loss cows. The serum concentration of leptin ap was greater in OC-loss than in the OC-no loss cows. Overall, OC cows lost more BFT than normal or lean cows. However, those OC cows with a smaller loss of BFT produced less milk than OC cows with greater losses.

Plasma concentrations of branched-chain amino acids differ with Holstein genetic strain in pasture-based dairy systems

In pasture-based systems, there are nutritional and climatic challenges exacerbated across lactation; thus, dairy cows require an enhanced adaptive capacity compared with cows in confined systems. We aimed to evaluate the effect of lactation stage (21 vs. 180 days in milk, DIM) and Holstein genetic strain (North American Holstein, NAH, n = 8; New Zealand Holstein, NZH, n = 8) on metabolic adaptations of grazing dairy cows through plasma metabolomic profiling and its association with classical metabolites. Although 67 metabolites were affected (FDR < 0.05) by DIM, no metabolite was observed to differ between genetic strains while only alanine was affected (FDR = 0.02) by the interaction between genetic strain and DIM. However, complementary tools for time-series analysis (ASCA analysis, MEBA ranking) indicated that alanine and the branched-chain amino acids (BCAA) differed between genetic strains in a lactation-stage dependent manner. Indeed, NZH cows had lower (P-Tukey < 0.05) plasma concentrations of leucine, isoleucine and valine than NAH cows at 21 DIM, probably signaling for greater insulin sensitivity. Metabolic pathway analysis also revealed that, independently of genetic strains, AA metabolism might be structurally involved in homeorhetic changes as 40% (19/46) of metabolic pathways differentially expressed (FDR < 0.05) between 21 and 180 DIM belonged to AA metabolism.

Longitudinal changes in fatty acid metabolism and in the mitochondrial protein import system in overconditioned and normal conditioned cows: A transcriptional study using microfluidic quantitative PCR

This study investigated the effect of body condition around calving on the hepatic mRNA expression of genes involved in fatty acid (FA) metabolism and mitochondrial protein import system of dairy cows during the transition period. Fifteen weeks before their anticipated calving date, 38 multiparous Holstein cows were selected based on their current and previous body condition scores (BCS) and allocated to either a high or a normal BCS group (19 cows each). They received different diets to reach targeted differences in BCS and backfat thickness (BFT) until dry-off. At dry-off, normal BCS (NBCS) cows had a BCS <3.5 and BFT <1.2 cm, and the high BCS (HBCS) cows had a BCS >3.75 and BFT >1.4 cm. The expression of targeted genes in the liver was assayed by reverse-transcription quantitative real-time PCR using microfluidics integrated fluidic circuit chips on a subset of 5 cows from each group. Liver biopsies were collected at d -49, +3, +21, and +84 relative to parturition. The mRNA abundance of 47 genes related to lipid metabolism including carnitine metabolism, FA uptake and transport, lipoprotein export, carnitine metabolism, mitochondrial and proximal FA oxidation, ketogenesis, AMP-activated protein kinase/mammalian target of rapamycin pathway, and mitochondrial protein import system was assessed in liver tissue. The mRNA abundances of FA binding protein (FABP)6 (in both groups), and FABP1 and solute carrier family 22 member 5 (SLC22A5) in HBCS were upregulated (>1.5-fold change, FC) in early lactation (at d +3 and +21 postpartum) compared with antepartum (d -49), indicating promoted FA uptake and intracellular transport in the liver due to the metabolic adaptations of elevated lipo-mobilization after parturition. The upregulation of SLC22A5 and SLC25A20 after parturition was more pronounced in HBCS than in NBCS cows, suggesting a need for increasing the capacity of FA uptake, and FA transport into the hepatocyte. The increased mRNA abundance of carnitine palmitoyltransferase 1A, after parturition and to a greater extent in HBCS (FC = 4.1) versus NBCS (FC = 2.1) indicates a physiological increase in the capacity of long-chain fatty acyl-CoA entry into the liver mitochondria compared with antepartum (ap; d -49 relative to calving). The greater hepatic mRNA abundance of genes encoding enzymes involved in mitochondrial FA oxidation in HBCS than in NBCS points to an increased rate of mitochondrial β-oxidation. The hepatic mRNA abundance of 3-hydroxy-3-methylglutaryl-CoA synthase 2 and 3-hydroxy-3-methylglutaryl-CoA were upregulated after parturition (d +21/d +3 pp) to a greater extent in HBCS than in NBCS cows, indicating that excess acetyl-CoA generated via β-oxidation was increasingly used for ketogenesis. We observed for the first time that the mRNA abundance of genes involved in the translocase of the inner membrane (TIM) complex (TIM22 and TIM23) in the hepatic mitochondrial protein import system were undergoing distinct changes during the transition from late pregnancy to early lactation in dairy cows. Even though sample size in this study was relatively small, the results support that overconditioning around calving may contribute to mitochondrial FA overload and greater ketogenesis at the level of transcription in the liver of early lactation cows.

Machine Learning in Agriculture: A Comprehensive Updated Review

The digital transformation of agriculture has evolved various aspects of management into artificial intelligent systems for the sake of making value from the ever-increasing data originated from numerous sources. A subset of artificial intelligence, namely machine learning, has a considerable potential to handle numerous challenges in the establishment of knowledge-based farming systems. The present study aims at shedding light on machine learning in agriculture by thoroughly reviewing the recent scholarly literature based on keywords' combinations of "machine learning" along with "crop management", "water management", "soil management", and "livestock management", and in accordance with PRISMA guidelines. Only journal papers were considered eligible that were published within 2018-2020. The results indicated that this topic pertains to different disciplines that favour convergence research at the international level. Furthermore, crop management was observed to be at the centre of attention. A plethora of machine learning algorithms were used, with those belonging to Artificial Neural Networks being more efficient. In addition, maize and wheat as well as cattle and sheep were the most investigated crops and animals, respectively. Finally, a variety of sensors, attached on satellites and unmanned ground and aerial vehicles, have been utilized as a means of getting reliable input data for the data analyses. It is anticipated that this study will constitute a beneficial guide to all stakeholders towards enhancing awareness of the potential advantages of using machine learning in agriculture and contributing to a more systematic research on this topic.

Muscle metabolome and adipose tissue mRNA expression of lipid metabolism-related genes in over-conditioned dairy cows differing in serum-metabotype

Over-conditioned dairy cows, classified by body condition score (BCS) and backfat thickness (BFT) are less able to metabolically adapt to the rapidly increasing milk yield after parturition. Based on serum metabolome and cluster analyses, high BCS cows (HBCS) could be classified into metabotypes that are more similar to normal (NBCS) cows, i.e., HBCS predicted normal (HBCS-PN) than the HBCS predicted high (HBCS-PH) cows—similar to the concept of obese but metabolically healthy humans. Our objective was to compare muscle metabolome and mRNA abundance of genes related to lipogenesis and lipolysis in adipose tissue between HBCS-PH (n = 13), HBCS-PN (n = 6), and NBCS-PN (n = 15). Tail-head subcutaneous fat was biopsied on d −49, 3, 21, and 84 relative to parturition. Potential differences in the oxidative capacity of skeletal muscle were assessed by targeted metabolomics in M. semitendinosus from d 21. Besides characteristic changes with time, differences in the mRNA abundance were limited to lipogenesis-related genes on d −49 (HBCS-PH > HBCS-PN). The HBCS-PH had more than two-fold higher muscle concentrations of short (C2, C4-OH, C6-OH) and long-chain acylcarnitines (C16, C18, and C18:1) than HBCS-PN, indicating a greater oxidative capacity for fatty acids (and utilization of ketones) in muscle of HBCS-PN than HBCS-PH cows.

Differences in net fat oxidation, heat production, and liver mitochondrial DNA copy numbers between high and low feed-efficient dairy cows

Improving feed utilization efficiency in dairy cattle could have positive economic and environmental effects that would support the sustainability of the dairy industry. Identifying key differences in metabolism between high and low feed-efficient animals is vital to enhancing feed conversion efficiency. Therefore, our objectives were (1) to determine whether cows grouped by either high or low feed efficiency have measurable differences in net fat and carbohydrate metabolism that account for differences in heat production (HP), and if so, whether these differences also exists under conditions of feed withdrawal when the effect of feeding on HP is minimized, and (2) to determine whether the abundance of mitochondria in the liver can be related to the high or low feed-efficient groups. Ten dairy cows from a herd of 15 (parity = 2) were retrospectively grouped into either a high (H) or a low (L) feed-efficient group (n = 5 per group) based on weekly energy-corrected milk (ECM) divided by dry mater intake (DMI) from wk 4 through 30 of lactation. Livers were biopsied at wk -4, 2, and 12, and blood was sampled weekly from wk -3 to 12 relative to parturition. Blood was subset to be analyzed for the transition period (wk -3 to 3) and from wk 4 to 12. In wk 5.70 ± 0.82 (mean ± SD) postpartum (PP), cows spent 2 d in respiration chambers (RC), in which CO₂, O₂, and CH₄ gases were measured every 6 min for 24 h. Fatty acid oxidation (FOX), carbohydrate oxidation (COX), metabolic respiratory quotient (RQ), and HP were calculated from gas measurements for 23 h. Cows were fed ad libitum (AD-LIB) on d 1 and had feed withdrawn (RES, restricted diet) on d 2. Additional blood samples were taken at the end of the AD-LIB and RES feeding periods in the RC. During wk 4 to 30 PP, H had greater DMI/kg of metabolic body weight (BW^0.75), ECM per kilogram of BW^0.75 yield, and ECM/DMI ratio, compared with L, but a lower body condition score between wk 4 and 12 PP. In the RC period, we detected no differences in BW, DMI, or milk yield between groups. We also detected no significant group or group by feeding period interactions for plasma metabolites except for Revised Quantitative Insulin Sensitivity Check Index, which tended to have a group by feeding period interaction. The H group had lower HP and HP per kilogram of BW^0.75 compared with L. Additionally, H had lower FOX and FOX per kilogram of BW^0.75 compared with L during the AD-LIB period. Methane, CH₄ per kilogram of BW^0.75, and CH₄ per kilogram of milk yield were lower in H compared with L, but, when adjusted for DMI, CH₄/DMI did not differ between groups, nor did HP/DMI. Relative mitochondrial DNA copy numbers in the liver were lower in the L than in the H group. These results suggest that lower feed efficiency in dairy cows may result from fewer mitochondria per liver cell as well as a greater whole-body HP, which likely partially results from higher net fat oxidation.