Genetic parameter estimates for methane emission from breath during lactation and potential inaccuracies in reliabilities assuming a repeatability versus random regression model.

Methane emissions will be added to many national ruminant breeding programs in the coming years. Little is known about the covariance structure of CH4 traits over a lactation, which is important for optimizing recording strategies and establishing optimal genetic evaluation models. Our aim was to study CH4 over a lactation using random regression (RR) models, and to compare the accuracy to a fixed regression repeatability model under different phenotyping strategies. Data were available from repeated measurements of CH4 concentrations (ppm) recorded in the feed bins of milking robots on 52 commercial dairy farms in the Netherlands. In total, 36,370 averaged weekly records were available from 4,664 cows. Genetic parameters were estimated using a fixed regression model, and a RR model with first- to fifth-order Legendre polynomials for the additive genetic and within-lactation permanent environmental effect. The mean heritability (± SE) was 0.17 ± 0.04, and the mean within-lactation repeatability was 0.56 ± 0.03. The genetic correlations between DIM were high and ranged from 0.34 ± 0.36 to 1.00 ± <0.01. Permanent environmental correlations showed large deviations and ranged from -0.73 ± 0.08 to 1.00 ± <0.01. With a large number of full lactation daughter CH4 records per bull, the reliability was not sensitive to using the fixed versus the RR model. However, when shorter periods were recorded at the start and end of the lactation, the fixed regression model resulted in a loss of reliability up to 28% for bulls. Assuming the fixed model when the true (co)variance structure is reflected by the RR model, more than twice as long of a recording from the start of lactation was required to achieve maximum reliability for a bull. Thus, a too simplistic model could result in implementing too little recording, and in lower genetic gains than predicted from the reliability.

Tracking individual broilers on video in terms of time and distance.

Tracking group-housed individual broilers using video can provide valuable information on their health, welfare, and performance, allowing breeders to identify novel or indicator traits that aid genetic improvement. However, their similar appearances make tracking individual broilers in a group-housed setting challenging. This study aimed to analyze broiler tracking on video (number of ID-switches, tracking time and distance) and examined potential tracking errors (ID-losses - location, proximity, kinematics) in an experimental pen to enable broiler locomotion phenotyping. This comprehensive analysis provided insights into the potential and challenges of tracking group-housed broilers on video with regards to phenotyping broiler locomotion. Thirty-nine broilers, of which 35 noncolor marked, were housed in an experimental pen (1.80 × 2.61 m), and only data at 18 d of age were used. A YOLOv7-tiny model was trained (n = 140), validated (n = 30), and tested (n = 30) on 200 annotated frames to detect the broilers. On the test set, YOLOv7-tiny had a precision, recall, and average precision (@0.5 - Intersection over Union threshold) of 0.99. A multi-object tracker (SORT) was implemented and evaluated on ground-truth trajectories of thirteen white broilers based on 136 min of video data (1-min intervals). The number of ID-switches varied from 5 to 20 (mean: 9.92) per ground-truth trajectory, tracking times ranged from 1 (by definition) to 51 min (mean: 12.36), and tracking distances ranged from 0.01 to 17.07 meters (mean: 1.89) per tracklet. Tracking errors primarily occurred when broilers were occluded by the drinker, and relatively frequently when broilers were in close proximity (within 10 cm), with velocity and acceleration appearing to have a lesser impact on tracking errors. The study establishes a 'baseline' for future research and identified the potential and challenges of tracking group-housed individual broilers. The results highlighted the importance of addressing ID-switches, identified potential tracking algorithm improvements, and emphasized the need for an external animal identification system to enable objective, simultaneous and semi-continuous locomotion phenotyping of group-housed individual broilers.

Novel genetic parameters for genetic residual feed intake in dairy cattle using time series data from multiple parities and countries in North America and Europe.

Residual feed intake is viewed as an important trait in breeding programs that could be used to enhance genetic progress in feed efficiency. In particular, improving feed efficiency could improve both economic and environmental sustainability in the dairy cattle industry. However, data remain sparse, limiting the development of reliable genomic evaluations across lactation and parity for residual feed intake. Here, we estimated novel genetic parameters for genetic residual feed intake (gRFI) across the first, second, and third parity, using a random regression model. Research data on the measured feed intake, milk production, and body weight of 7,379 cows (271,080 records) from 6 countries in 2 continents were shared through the Horizon 2020 project Genomic Management Tools to Optimise Resilience and Efficiency, and the Resilient Dairy Genome Project. The countries included Canada (1,053 cows with 47,130 weekly records), Denmark (1,045 cows with 72,760 weekly records), France (329 cows with 16,888 weekly records), Germany (938 cows with 32,614 weekly records), the Netherlands (2,051 cows with 57,830 weekly records), and United States (1,963 cows with 43,858 weekly records). Each trait had variance components estimated from first to third parity, using a random regression model across countries. Genetic residual feed intake was found to be heritable in all 3 parities, with first parity being predominant (range: 22-34%). Genetic residual feed intake was highly correlated across parities for mid- to late lactation; however, genetic correlation across parities was lower during early lactation, especially when comparing first and third parity. We estimated a genetic correlation of 0.77 ± 0.37 between North America and Europe for dry matter intake at first parity. Published literature on genetic correlations between high input countries/continents for dry matter intake support a high genetic correlation for dry matter intake. In conclusion, our results demonstrate the feasibility of estimating variance components for gRFI across parities, and the value of sharing data on scarce phenotypes across countries. These results can potentially be implemented in genetic evaluations for gRFI in dairy cattle.

Heritability and genetic correlations between enteric methane production and concentration recorded by GreenFeed and sniffers on dairy cows.

To reduce methane (CH4) emissions of dairy cows by animal breeding, CH4 measurements have to be recorded on thousands of individual cows. Currently, several techniques are used to phenotype cows for CH4, differing in costs and applicability. However, there is uncertainty about the agreement between techniques. To judge the similarity and repeatability between measurements of different recording techniques, the repeatability, heritability, and genetic correlation are useful metrics. Therefore, our objective was to estimate (1) the repeatability and heritability for CH4 and carbon dioxide production recorded by GreenFeed (GF) and for CH4 and carbon dioxide concentration measured by cost-effective but less accurate sniffers, and (2) the genetic correlation between CH4 recorded with these 2 different on farm and high throughput techniques. Data were available from repeated measurements of CH4 production (grams/day) by GF units and of CH4 concentration (ppm) by sniffers, recorded on commercial dairy farms in the Netherlands. The final data comprised 24,284 GF daily means from 822 cows, 170,826 sniffer daily means from 1,800 cows, and 1,786 daily means from 75 cows by both GF and sniffer (in the same period). Additionally, CH4 records were averaged per week. For daily and weekly mean GF CH4 the heritabilities were 0.19 ± 0.02 and 0.33 ± 0.04, and for daily and weekly mean sniffer CH4 the heritabilities were similar and were 0.18 ± 0.01 and 0.32 ± 0.02, respectively. Phenotypic correlations between GF CH4 production and sniffer CH4 concentration were moderate (0.39 ± 0.03 for daily means and 0.37 ± 0.05 for weekly means). However, genetic correlations were high; 0.71 ± 0.13 for daily means and 0.76 ± 0.15 for weekly means. The high genetic correlation indicates that selection on low CH4 concentrations (ppm) recorded by the cost-effective sniffer method, will result in reduced CH4 production (grams/day) as recorded with GF.

Passive radio frequency identification and video tracking for the determination of location and movement of broilers.

Phenotypes on individual animals are required for breeding programs to be able to select for traits. However, phenotyping individual animals can be difficult and time-consuming, especially for traits related to health, welfare, and performance. Individual broiler behavior could serve as a proxy for these traits when recorded automatically and reliably on many animals. Sensors could record individual broiler behavior, yet different sensors can differ in their assessment. In this study a comparison was made between a passive radio frequency identification (RFID) system (grid of antennas underneath the pen) and video tracking for the determination of location and movement of 3 color-marked broilers at d 18. Furthermore, a systems comparison of derived behavioral metrics such as space usage, locomotion activity and apparent feeding and drinking behavior was made. Color-marked broilers simplified the computer vision task for YOLOv5 to detect, track, and identify the animals. Animal locations derived from the RFID-system and based on video were largely in agreement. Most location differences (77.5%) were within the mean radius of the antennas' enclosing circle (≤128 px, 28.15 cm), and 95.3% of the differences were within a one antenna difference (≤256 px, 56.30 cm). Animal movement was not always registered by the RFID-system whereas video was sensitive to detection noise and the animal's behavior (e.g., pecking). The method used to determine location and the systems' sensitivities to movement led to differences in behavioral metrics. Behavioral metrics derived from video are likely more accurate than RFID-system derived behavioral metrics. However, at present, only the RFID-system can provide individual identification for non-color marked broilers. A combination of verifiable and detailed video with the unique identification of RFID could make it possible to identify, describe, and quantify a wide range of individual broiler behaviors.

Rapid turnover of sensor data to genetic evaluation for dairy cows in the cloud.

More and more sensor and automation data are available that enable animal breeders to define novel traits. However, sensor and automation data are often frequently measured differently (e.g., milk yield and different milk components are continuously measured during each milking). These differences are challenging animal breeders to define traits and use the most appropriate analytical models for genetic evaluation and breeding values. Traditionally, the process from raw data to breeding value estimations involves several steps: data curation, trait definition, variance component estimation, genetic evaluation, and validation of the estimated breeding values (EBV). All these steps often take many iterations and several research projects to optimize the final genetic evaluations. To make this entire process-from raw data to validated EBV-more efficient, we combined all these steps in a cloud environment that allows for faster processing and a faster data distribution time. We used real data (including 1,782,373,113 daily milk-yield records of 1,120,550 dairy cows) and a real trait (a resilience trait based on the deviations from expected milk yields) to demonstrate the functioning of this cloud environment. The daily milk-yield records were incorporated into our cloud solution, in which we have set up central binary large object storage. Subsequent steps were all performed in the cloud. The data set was preprocessed in approximately 6 h to obtain the resilience indicator for 352,871 cows in the first 3 lactations. Estimation of genetic parameters (heritabilities and genetic correlations) was performed by splitting the data into 5 subsets in ASReml, and prediction of subsequent EBV was performed on the entire data set using MiXBLUP. Together with the validation of breeding values, this process encompassed 16.5 h. By combining the different steps from preprocessing sensor data to genetic evaluation of new traits in one cloud environment, we generated EBV and validation plots in approximately 1 working day. Moreover, our setup is a flexible design and can be adapted easily to test new, longitudinal sensor-driven traits and compare the performance of these new traits to previous ones.

Observational study on associations between resilience indicators based on daily milk yield in first lactation and lifetime profitability.

Resilience is the ability of cows to be minimally affected by disturbances, such as pathogens, heat waves, and changes in feed quality, or to quickly recover. Obvious advantages of resilience are good animal welfare and easy and pleasant management for farmers. Furthermore, economic effects are also expected, but these remain to be determined. The goal of this study was to investigate the association between resilience and lifetime gross margin, using indicators of resilience calculated from fluctuations in daily milk yield using an observational study. Resilience indicators and lifetime gross margin were calculated for 1,325 cows from 21 herds. These cows were not alive anymore and, therefore, had complete lifetime data available for many traits. The resilience indicators were the natural log-transformed variance (LnVar) and the lag-1 autocorrelation (rauto) of daily milk yield deviations from cow-specific lactation curves in parity 1. Good resilience is indicated by low LnVar (small yield response to disturbances) and low rauto (quick yield recovery to baseline). Lifetime gross margin was calculated as the sum of all revenues minus the sum of all costs throughout life. Included revenues were from milk, calf value, and slaughter of the cow. Included costs were from feed, rearing, insemination, management around calving, disease treatments, and destruction in case of death on farm. Feed intake was unknown and, therefore, lifetime feed costs had to be estimated based on milk yield records. The association of each resilience indicator with lifetime gross margin, and also with the underlying revenues and costs, was investigated using analysis of covariance (ANCOVA) models. Mean daily milk yield in first lactation, herd, and year of birth were included as covariates and factors. Natural log-transformed variance had a significantly negative association with lifetime gross margin, which means that cows with stable milk yield (low LnVar, good resilience) in parity 1 generated on average a higher lifetime gross margin than cows that had the same milk yield level but with more fluctuations. The association with lifetime gross margin could be mainly attributed to higher lifetime milk revenues for cows with low LnVar, due to a longer lifespan. Unlike LnVar, rauto was not significantly associated with lifetime gross margin or any of the underlying lifetime costs and revenues. However, it was significantly associated with yearly treatment costs, which is important for ease of management. In conclusion, the importance of resilience for total profit generated by a cow at the end of life was confirmed by the significant association of LnVar with lifetime gross margin, although effects of differences in feed efficiency between resilient and less resilient cows remain to be studied. The economic advantage can be mainly ascribed to benefits of long lifespan.

Genetic parameters for repeatedly recorded enteric methane concentrations of dairy cows.

Animal breeding techniques offer potential to reduce enteric emissions of ruminants to lower the environmental impact of dairy farming. The aim of this study was to estimate the heritability and repeatability of methane (CH4) concentrations, using the largest data set from long-term repeatedly recorded CH4 on cows to date, and to evaluate (1) the accuracy of breeding values for different CH4 traits, including using visits or weekly means, and (2) recording strategies (with varying numbers of records and recorded daughters per sire). The data comprised of long-term recording of CH4 and carbon dioxide (CO2), from 1,746 Holstein Friesian cows, on 14 commercial dairy farms throughout the Netherlands. Emissions were recorded in 10- to 35-s intervals, between 64 and 436 d, depending on farms. From each robot visit, CH4 and CO2 concentrations were summarized into various traits, averaged per visit and per week: mean, median, mean log, and mean CH4/CO2 ratio. Genetic parameters were estimated with animal repeatability models, using a restricted maximum likelihood procedure, and a relationship matrix based on genotypes and pedigree. The heritability was equal for mean and median CH4 per visit (0.13) but lower for logCH4 and CH4/CO2 (0.07 and 0.01, respectively). Phenotypic and genetic correlations were high (≥0.78) between the CH4 traits, apart from the genetic correlations with the CH4/CO2 trait, which were negative. To achieve a minimum reliability of 50% for the estimated breeding value of a bull, 25 records on mean CH4, measured on 10 different daughters, were sufficient. Although the heritability and repeatability were higher for weekly (0.32 and 0.68, respectively) than for visit mean CH4 (0.13 and 0.30, respectively), the reliabilities of estimated breeding values from visit or weekly means were equal; thus, we found no advantage in averaging records to weekly means for genetic evaluations.

Selective breeding as a mitigation tool for methane emissions from dairy cattle.

The global livestock sector, particularly ruminants, contributes substantially to the total anthropogenic greenhouse gases. Management and dietary solutions to reduce enteric methane (CH4) emissions are extensively researched. Animal breeding that exploits natural variation in CH4 emissions is an additional mitigation solution that is cost-effective, permanent, and cumulative. We quantified the effect of including CH4 production in the Dutch breeding goal using selection index theory. The current Dutch national index contains 15 traits, related to milk yield, longevity, health, fertility, conformation and feed efficiency. From the literature, we obtained a heritability of 0.21 for enteric CH4 production, and genetic correlations of 0.4 with milk lactose, protein, fat and DM intake. Correlations between enteric CH4 production and other traits in the breeding goal were set to zero. When including CH4 production in the current breeding goal with a zero economic value, CH4 production increases each year by 1.5 g/d as a correlated response. When extrapolating this, the average daily CH4 production of 392 g/d in 2018 will increase to 442 g/d in 2050 (+13%). However, expressing the CH4 production as CH4 intensity in the same period shows a reduction of 13%. By putting economic weight on CH4 production in the breeding goal, selective breeding can reduce the CH4 intensity even by 24% in 2050. This shows that breeding is a valuable contribution to the whole set of mitigation strategies that could be applied in order to achieve the goals for 2050 set by the EU. If the decision is made to implement animal breeding strategies to reduce enteric CH4 production, and to achieve the expected breeding impact, there needs to be a sufficient reliability of prediction. The only way to achieve that is to have enough animals phenotyped and genotyped. The power calculations offer insights into the difficulties that will be faced in trying to record enough data. Recording CH4 data on 100 farms (with on average 150 cows each) for at least 2 years is required to achieve the desired reliability of 0.40 for the genomic prediction.

Validation of resilience indicators by estimating genetic correlations among daughter groups and with yield responses to a heat wave and disturbances at herd level.

Resilient cows are minimally affected in their functioning by disturbances, and if affected, they quickly recover. Previously, the variance and autocorrelation of daily deviations from a lactation curve were proposed as resilience indicators. These traits were heritable and genetically associated with good health and longevity. However, it was unknown if selection for these indicators would lead to desired changes in the phenotype. The first aim of this study was to investigate if forward prediction of the resilience indicators in another environment was possible. Therefore, the resilience indicator records were split into 2 subsets, each containing half of the daughters of each sire, split within sire into cows that calved in early year-seasons and cows that calved in more recent year-seasons. Genetic correlations between the subsets were then estimated for each resilience indicator. The second aim was to estimate genetic correlations between the resilience indicators and traits describing production responses to actual disturbances. The disturbances were a heat wave in July 2015 and yield disturbances at herd level. The latter were selected by decreases in mean yield of all primiparous cows in a herd, indicating that a disturbance occurred. The data set used for calculation of the resilience indicators and the traits describing yield responses contained 62,932,794 daily milk yield records on 199,104 primiparous cows. Genetic correlations (rg) between recent and earlier daughter groups were 1 for both resilience indicators, which suggests that selection will result in changes in the phenotype in the next generation. Furthermore, low variance was genetically correlated with weak response in milk yield to both the heat wave and herd disturbances (rg 0.47 to 0.97). Low autocorrelation was genetically correlated with reduced perturbation length and quick recovery after the heat wave and herd disturbances (0.28 to 0.97). These results suggest that variance and autocorrelation cover different aspects of resilience, and should be combined in a resilience index. In conclusion, genetic selection for the resilience indicators will likely result in favorable changes in the traits themselves, and in response and recovery to actual disturbances, which confirms that they are useful resilience indicators.

Between-herd variation in resilience and relations to herd performance.

Resilient cows are minimally affected in their functioning by infections and other disturbances, and recover quickly. Herd management is expected to have an effect on disturbances and the resilience of cows, and this effect was investigated in this study. Two resilience indicators were first recorded on individual cows. The effect of herd-year on these resilience indicators was then estimated and corrected for genetic and year-season effects. The 2 resilience indicators were the variance and the lag-1 autocorrelation of daily milk yield deviations from an expected lactation curve. Low variance and autocorrelation indicate that a cow does not fluctuate much around her expected milk yield and is, thus, subject to few disturbances, or little affected by disturbances (resilient). The herd-year estimates of the resilience indicators were estimated for 9,917 herd-year classes based on records of 227,655 primiparous cows from 2,644 herds. The herd-year estimates of the resilience indicators were then related to herd performance variables. Large differences in the herd-year estimates of the 2 resilience indicators (variance and autocorrelation) were observed between herd-years, indicating an effect of management on these traits. Furthermore, herd-year classes with a high variance tended to have a high proportion of cows with a rumen acidosis indication (r = 0.31), high SCS (r = 0.19), low fat content (r = -0.18), long calving interval (r = 0.14), low survival to second lactation (r = -0.13), large herd size (r = 0.12), low lactose content (r = -0.12), and high production (r = 0.10). These correlations support that herds with high variance are not resilient. The correlation between the variance and the proportion of cows with a rumen acidosis indication suggests that feed management may have an important effect on the variance. Herd-year classes with a high autocorrelation tended to have a high proportion of cows with a ketosis indication (r = 0.14) and a high production (r = 0.13), but a low somatic cell score (r = -0.17) and a low proportion of cows with a rumen acidosis indication (r = -0.12). These correlations suggest that high autocorrelation at herd level indicates either good or poor resilience, and is thus a poor resilience indicator. However, the combination of a high variance and a high autocorrelation is expected to indicate many fluctuations with slow recovery. In conclusion, herd management, in particular feed management, seems to affect herd resilience.

Storing, combining and analysing turkey experimental data in the Big Data era.

With the increasing availability of large amounts of data in the livestock domain, we face the challenge to store, combine and analyse these data efficiently. With this study, we explored the use of a data lake for storing and analysing data to improve scalability and interoperability. Data originated from a 2-day animal experiment in which the gait score of approximately 200 turkeys was determined through visual inspection by an expert. Additionally, inertial measurement units (IMUs), a 3D-video camera and a force plate (FP) were installed to explore the effectiveness of these sensors in automating the visual gait scoring. We deployed a data lake using the IMU and FP data of a single day of that animal experiment. This encompasses data from 84 turkeys for which we preprocessed by performing an 'extract, transform and load' (ETL-) procedure. To test scalability of the ETL-procedure, we simulated increasing volumes of the available data from this animal experiment and computed the 'wall time' (elapsed real time) for converting FP data into comma-separated files and storing these files. With a simulated data set of 30 000 turkeys, the wall time reduced from 1 h to less than 15 min, when 12 cores were used compared to 1 core. This demonstrated the ETL-procedure to be scalable. Subsequently, a machine learning (ML) pipeline was developed to test the potential of a data lake to automatically distinguish between two classses, that is, very bad gait scores v. other scores. In conclusion, we have set up a dedicated customized data lake, loaded data and developed a prediction model via the creation of an ML pipeline. A data lake appears to be a useful tool to face the challenge of storing, combining and analysing increasing volumes of data of varying nature in an effective manner.

Can greenhouse gases in breath be used to genetically improve feed efficiency of dairy cows?

There is considerable interest in improving feed utilization of dairy cattle while limiting losses to the environment (i.e., greenhouse gases, GHG). To breed for feed-efficient or climate-friendly cattle, it is first necessary to obtain accurate estimates of genetic parameters and correlations of feed intake, greenhouse gases, and production traits. Reducing dry matter take (DMI) requirements while maintaining production has high economic value to farmers, but DMI is costly to record and thus limited to small research or nucleus herds. Conversely, enteric methane (CH4) currently has no economic value, is also costly to record, and is limited to small experimental trials. However, breath gas concentrations of methane (CH4c) and carbon dioxide (CO2c) are relatively cheap to measure at high throughput under commercial conditions by installing sniffers in automated milking stations. The objective of this study was to assess the genetic correlations between DMI, body weight (BW), fat- and protein-corrected milk yield (FPCM), and GHG-related traits: CH4c and CO2c from Denmark (DNK) and the Netherlands (NLD). A second objective was to assess the genetic potential for improving feed efficiency and the added benefits of using CH4c and CO2c as indicators. Feed intake data were available on 703 primiparous cows in DNK and 524 in NLD; CH4c and CO2c records were available on 434 primiparous cows in DNK and 656 in NLD. The GHG-related traits were heritable (e.g., CH4c h2: DNK = 0.26, NLD = 0.15) but were differentially genetically correlated with DMI and feed efficiency in both magnitude and sign, depending on the population and the definition of feed efficiency. Across feed efficiency traits and DMI, having bulls with 100 daughters with FPCM, BW, and GHG traits resulted in sufficiently high accuracy to almost negate the need for DMI records. Despite differences in genetic correlation structure, the relatively cheap GHG-related traits showed considerable potential for improving the accuracy of breeding values of highly valuable feed intake and feed efficiency traits.

Predicting survival in dairy cattle by combining genomic breeding values and phenotypic information.

Advances in technology and improved data collection have increased the availability of genomic estimated breeding values (gEBV) and phenotypic information on dairy farms. This information could be used for the prediction of complex traits such as survival, which can in turn be used in replacement heifer management. In this study, we investigated which gEBV and phenotypic variables are of use in the prediction of survival. Survival was defined as survival to second lactation, plus 2 wk, a binary trait. A data set was obtained of 6,847 heifers that were all genotyped at birth. Each heifer had 50 gEBV and up to 62 phenotypic variables that became gradually available over time. Stepwise variable selection on 70% of the data was used to create multiple regression models to predict survival with data available at 5 decision moments: distinct points in the life of a heifer at which new phenotypic information becomes available. The remaining 30% of the data were kept apart to investigate predictive performance of the models on independent data. A combination of gEBV and phenotypic variables always resulted in the model with the highest Akaike information criterion value. The gEBV selected were longevity, feet and leg score, exterior score, udder score, and udder health score. Phenotypic variables on fertility, age at first calving, and milk quantity were important once available. It was impossible to predict individual survival accurately, but the mean predicted probability of survival of the surviving heifers was always higher than the mean predicted probability of the nonsurviving group (difference ranged from 0.014 to 0.028). The model obtained 2.0 to 3.0% more surviving heifers when the highest scoring 50% of heifers were selected compared with randomly selected heifers. Combining phenotypic information and gEBV always resulted in the highest scoring models for the prediction of survival, and especially improved early predictive performance. By selecting the heifers with the highest predicted probability of survival, increased survival could be realized at the population level in practice.

Exploration of variance, autocorrelation, and skewness of deviations from lactation curves as resilience indicators for breeding.

The ability of a cow to cope with environmental disturbances, such as pathogens or heat waves, is called resilience. To improve resilience through breeding, we need resilience indicators, which could be based on the fluctuation patterns in milk yield resulting from disturbances. The aim of this study was to explore 3 traits that describe fluctuations in milk yield as indicators for breeding resilient cows: the variance, autocorrelation, and skewness of the deviations from individual lactation curves. We used daily milk yield records of 198,754 first-parity cows, recorded by automatic milking systems. First, we estimated a lactation curve for each cow using 4 different methods: moving average, moving median, quantile regression, and Wilmink curve. We then calculated the log-transformed variance (LnVar), lag-1 autocorrelation (rauto), and skewness (Skew) of the daily deviations from these curves as resilience indicators. A genetic analysis of the resilience indicators was performed, and genetic correlations between resilience indicators and health, longevity, fertility, metabolic, and production traits were estimated. The heritabilities differed between LnVar (0.20 to 0.24), rauto (0.08 to 0.10), and Skew (0.01 to 0.02), and the genetic correlations among the indicators were weak to moderate. For rauto and Skew, genetic correlations with health, longevity, fertility, and metabolic traits were weak or the opposite of what we expected. Therefore, rauto and Skew have limited value as resilience indicators. However, lower LnVar was genetically associated with better udder health (genetic correlations from -0.22 to -0.32), better longevity (-0.28 to -0.34), less ketosis (-0.27 to -0.33), better fertility (-0.06 to -0.17), higher BCS (-0.29 to -0.40), and greater dry matter intake (-0.53 to -0.66) at the same level of milk yield. These correlations support LnVar as an indicator of resilience. Of all 4 curve-fitting methods, LnVar based on quantile regression systematically had the strongest genetic correlations with health, longevity, and fertility traits. Thus, quantile regression is considered the best curve-fitting method. In conclusion, LnVar based on deviations from a quantile regression curve is a promising resilience indicator that can be used to breed cows that are better at coping with disturbances.

Genetic parameters for atypical reproductive patterns in dairy cows estimated from in-line milk progesterone profiles.

Our aim was to estimate genetic parameters of atypical reproductive patterns and estimate their genetic correlation with milk production and classical fertility traits for commercial dairy cows. In contrast with classical fertility traits, atypical reproductive patterns based on in-line milk progesterone profiles might have higher heritability and lower genetic correlation with milk production. We had in-line milk progesterone profiles available for 12,046 cycles in 4,170 lactations of 2,589 primiparous and multiparous cows (mainly Holstein Friesian) from 14 herds. Based on progesterone profiles, 5 types of atypical reproductive patterns in a lactation were defined: delayed ovulation types I and II, persistent corpus luteum types I and II, and late embryo mortality. These atypical patterns were detected in 14% (persistent corpus luteum type II) to 21% (persistent corpus luteum type I) of lactations. In 47% of lactations, at least 1 atypical pattern was detected. Threshold model heritabilities for atypical reproduction patterns ranged between 0.03 and 0.14 and for most traits were slightly higher compared with classical fertility traits. The genetic correlation between milk yield and calving interval was 0.56, whereas genetic correlations between milk yield and atypical reproductive patterns ranged between -0.02 and 0.33. Although most of these correlations between milk yield and atypical reproductive patterns are still unfavorable, they are lower compared with the correlations between classical fertility traits and milk yield. Therefore selection against atypical reproductive patterns may relax some constraints in current dairy breeding programs, to enhance genetic progress in both fertility and milk yield at a steady pace. However, as long as the target trait for fertility is calving interval, atypical reproductive patterns will not add additional value to the breeding goal in the near future due to the low number of available records.

Comparing regression, naive Bayes, and random forest methods in the prediction of individual survival to second lactation in Holstein cattle.

In this study, we compared multiple logistic regression, a linear method, to naive Bayes and random forest, 2 nonlinear machine-learning methods. We used all 3 methods to predict individual survival to second lactation in dairy heifers. The data set used for prediction contained 6,847 heifers born between January 2012 and June 2013, and had known survival outcomes. Each animal had 50 genomic estimated breeding values available at birth and up to 65 phenotypic variables that accumulated over time. Survival was predicted at 5 moments in life: at birth, at 18 mo, at first calving, at 6 wk after first calving, and at 200 d after first calving. The data sets were randomly split into 70% training and 30% testing sets to evaluate model performance for 20-fold validation. The methods were compared for accuracy, sensitivity, specificity, area under the curve (AUC) value, contrasts between groups for the prediction outcomes, and increase in surviving animals in a practical scenario. At birth and 18 mo, all methods had overlapping performance; no method significantly outperformed the other. At first calving, 6 wk after first calving, and 200 d after first calving, random forest and naive Bayes had overlapping performance, and both machine-learning methods outperformed multiple logistic regression. Overall, naive Bayes has the highest average AUC at all decision points up to 200 d after first calving. Random forest had the highest AUC at 200 d after first calving. All methods obtained similar increases in survival in the practical scenario. Despite this, the methods appeared to predict the survival of individual heifers differently. All methods improved over time, but the changes in mean model outcomes for surviving and non-surviving animals differed by method. Furthermore, the correlations of individual predictions between methods ranged from r = 0.417 to r = 0.700; the lowest correlations were at first calving for all methods. In short, all 3 methods were able to predict survival at a population level, because all methods improved survival in a practical scenario. However, depending on the method used, predictions for individual animals were quite different between methods.

Invited review: Determination of large-scale individual dry matter intake phenotypes in dairy cattle.

Feed efficiency has been widely studied in many areas of dairy science and is currently seeing renewed interest in the field of breeding and genetics. A critical part of determining how efficiently an animal utilizes feed is accurately measuring individual dry matter (DM) intake. Currently, multiple methods are used to measure feed intake or determine the DM content of that feed, resulting in different levels of accuracy of measurement. Furthermore, the scale at which data need to be collected for use in genetic analyses makes some methodologies impractical. This systematic review aims to provide an overview of the current methodologies used to measure both feed intake in ruminants and DM content of feedstuffs, current methods to predict individual DM intake, and applications of large-scale intake measurements. Overall, advances in milk spectral data analysis present a promising method of estimating individual DM intake on a herd scale with further validation of prediction models. Although measurements of individual feed intake rely on the same underlying principle, the methods selected are largely dictated by the costs of capital, labor, and necessary analyses. Finally, DM methodologies were synthesized into a comprehensive protocol for use in a variety of feedstuffs.

Value of the Dutch Holstein Friesian germplasm collection to increase genetic variability and improve genetic merit.

National gene bank collections for Holstein Friesian (HF) dairy cattle were set up in the 1990s. In this study, we assessed the value of bulls from the Dutch HF germplasm collection, also known as cryobank bulls, to increase genetic variability and improve genetic merit in the current bull population (bulls born in 2010-2015). Genetic variability was defined as 1 minus the mean genomic similarity (SIMSNP) or as 1 minus the mean pedigree-based kinship (fPED). Genetic merit was defined as the mean estimated breeding value for the total merit index or for 1 of 3 subindices (yield, fertility, and udder health). Using optimal contribution selection, we minimized relatedness (maximized variability) or maximized genetic merit at restricted levels of relatedness. We compared breeding schemes with only bulls from 2010 to 2015 with schemes in which cryobank bulls were also included. When we minimized relatedness, inclusion of genotyped cryobank bulls decreased mean SIMSNP by 0.7% and inclusion of both genotyped and nongenotyped cryobank bulls decreased mean fPED by 2.6% (in absolute terms). When we maximized merit at restricted levels of relatedness, inclusion of cryobank bulls provided additional merit at any level of mean SIMSNP or mean fPED except for the total merit index at high levels of mean SIMSNP. Additional merit from cryobank bulls depended on (1) the relative emphasis on genetic variability and (2) the selection criterion. Additional merit was higher when more emphasis was put on genetic variability. For fertility, for example, it was 1.74 SD at a mean SIMSNP restriction of 64.5% and 0.37 SD at a mean SIMSNP restriction of 67.5%. Additional merit was low to nonexistent for the total merit index and higher for the subindices, especially for fertility. At a mean SIMSNP of 64.5%, for example, it was 0.60 SD for the total merit index and 1.74 SD for fertility. In conclusion, Dutch HF cryobank bulls can be used to increase genetic variability and improve genetic merit in the current population, although their value is very limited when selecting for the current total merit index. Anticipating changes in the breeding goal in the future, the germplasm collection is a valuable resource for commercial breeding populations.

Improving accuracy of bulls' predicted genomic breeding values for fertility using daughters' milk progesterone profiles.

The main objective of this study was to investigate the benefit of accuracy of genomic prediction when combining records for an intermediate physiological phenotype in a training population with records for a traditional phenotype. Fertility was used as a case study, where commencement of luteal activity (C-LA) was the physiological phenotype, whereas the interval from calving to first service and calving interval were the traditional phenotypes. The potential accuracy of across-country genomic prediction and optimal recording strategies of C-LA were also investigated in terms of the number of farms and number of repeated records for C-LA. Predicted accuracy was obtained by estimating population parameters for the traits in a data set of 3,136 Holstein Friesian cows with 8,080 lactations and using a deterministic prediction equation. The effect of genetic correlation, heritability, and reliability of C-LA on the accuracy of genomic prediction were investigated. When the existing training population was 10,000 bulls with reliable estimated breeding value for the traditional trait, predicted accuracy for the physiological trait increased from 0.22 to 0.57 when 15,000 cows with C-LA records were added to the bull training population; but, when the interest was in predicting the traditional trait, we found no benefit from the additional recording. When the genetic correlation was higher between the physiological and traditional traits (0.7 instead of 0.3), accuracy increased less when adding the 15.000 cows with C-LA (from 0.51 to 0.63). In across-country predictions, we observed little to no increase in accuracy of the intermediate physiological phenotype when the training population from Sweden was large, but when accuracy increased the training population was small (200 cows), from 0.19 to 0.31 when 15,000 cows were added from the Netherlands (genetic correlation of 0.5 between countries), and from 0.19 to 0.48 for genetic correlation of 0.9. The predicted accuracy initially increased substantially when recording on the same farm was extended and multiple C-LA records per cow were used in prediction compared with single records; that is, accuracy increased from 0.33 with single records to 0.38 with multiple records (on average 1.6 records per cow) from 2 yr of recording C-LA. But, when the number C-LA per cow increased beyond 2 yr of recording, we noted no substantial benefit in accuracy from multiple records. For example, for 5 yr of recording (on average 2.5 records per cow), accuracy was 0.47; on doubling the recording period to 10 yr (on average 3.1 records per cow), accuracy increased by 0.07 units, whereas when C-LA was recorded for 15 yr (on average 3.3 records per cow) accuracy increased only by 0.05 units. Therefore, for genomic prediction using expensive equipment to record traits for training populations, it is important to optimize the recording strategy. The focus should be on recording more cows rather than continuous recording on the same cows.