Journal Article 0368 Genetic evaluation of dairy cow livability Get access J. R. Wright, J. R. Wright 1Animal Genomics and Improvement Laboratory, USDA-ARS, Beltsville, MD Search for other works by this author on: Oxford Academic PubMed Google Scholar P. M. VanRaden P. M. VanRaden 1Animal Genomics and Improvement Laboratory, USDA-ARS, Beltsville, MD Search for other works by this author on: Oxford Academic PubMed Google Scholar Journal of Animal Science, Volume 94, Issue suppl_5, October 2016, Page 178, https://doi.org/10.2527/jam2016-0368 Published: 01 October 2016

Flexible software was designed to replace the current animal model programs used for national genetic evaluations. Model improvements included (1) multi-trait processing, (2) multiple fixed class and regression variables, (3) differing models for different traits, (4) random regressions, and (5) foreign data included using pseudo-records. Computational improvements included (6) parallel processing, (7) renumbering class variables to equation numbers within the program so that estimated effects are output with original identification numbers, and (8) reliability computed within the same program. When applied to 3 fertility traits of 27,971,895 cows and heifers, the new model used daughter pregnancy rate as a correlated trait to improve heifer and cow conception rate evaluations for older animals and in herd-years where records are missing, and also added information from crossbreds. When applied to 7 traits and 76,846,327 lactation records of 30,064,300 cows, gains in accuracy were small for yield and somatic cell score, moderate for daughter pregnancy rate, and larger for productive life for recent bulls compared with single-trait evaluations. For very old bulls, multi-trait gains were also large for protein because lactation records were available only for milk and fat. Multi-trait productive life was computed with exact rather than approximate methods; however, correlated information from conformation was excluded, reducing advantages of the new model over the previous software. Estimates of breed differences, inbreeding depression, and heterosis were similar to previous estimates; new estimates were obtained for conception rates. Predictions were compared by truncating 4 yr of data, and genetic trend validation was applied to all breed-trait combinations. The estimates of trend account for increases in inbreeding across time. Incorporation of foreign data gave correlations above 0.98 for new with previous evaluations of foreign Holstein bulls, but lower for other breeds. The 7-trait model required 35 GB of memory and 3 d to converge using 7 processors. The new software was implemented for fertility traits in 2013 and is scheduled for implementation with yield, somatic cell score, and productive life in 2014. Further revision of the models and software may be needed in the near future to account for genomic preselection.

Potential biases from genomic pre-selection were estimated from actual selection and mating patterns of US Holsteins. Traditional models using only phenotypes and pedigrees do not adjust for average genomic merit of an animal’s parents, progeny, mates, or contemporaries. Positive assortative mating of elite young bulls to elite genotyped females and dams of highly selected sons will become primary sources of bias in the next few years unless methods of adjustment are introduced. However, deregression can remove some biases such as by crediting the dam for only her own records and not for those of her selected progeny. Diagonals of genomic and pedigree relationship matrices, their inverses, and their differences were examined. The animal’s own genotype provides a fourth source of information along with traditional information from parents, progeny, and own phenotype.

Genetic parameters were estimated for mobility score and 16 linear type traits of Brown Swiss dairy cattle. Mobility is an overall assessment trait that measures a cow's ability to move, as well as the structure of her feet, pasterns, and legs. Scores from 50 to 99 were assigned by appraisers for the Brown Swiss Cattle Breeders' Association of the USA beginning in June 2007. Only scores made before 69 mo of age were used. After edits, 32,710 records were available for 19,472 cows in 819 herds. The model included fixed effects for the interaction of herd and appraisal date (2,109 groups), appraisal age within parity (46 groups), and lactation stage within parity (21 groups), as well as random effects for animal, permanent environment, and residual error. A multi-trait analysis was conducted using canonical transformation, multiple diagonalization, and a decelerated expectation-maximization REML algorithm. Heritability was estimated to be 0.21 for mobility and ranged from 0.06 to 0.37 for the other 16 type traits. The traits with the highest genetic correlation with mobility were final score (0.78), rear legs (rear view; 0.74), rear udder width (0.52), and foot angle (0.51). Predicted transmitting ability (PTA) for mobility was calculated using the Brown Swiss multi-trait type evaluation system but included only appraisals for which all traits had been scored. For the 1,868 bulls evaluated, PTA for mobility ranged from 1.6 to -1.8 with a standard deviation of 0.5 and was most highly correlated with PTA for final score (0.88), rear legs (rear view; 0.77), rear udder height (0.70), and rear udder width (0.69), as expected from the trait correlations. When matched with official US national evaluations from August 2011, PTA mobility had moderately high correlations with PTA for milk, fat, and protein yields, as well as productive life (0.31-0.41). The mobility trait may provide a more accurate assessment of the structural soundness required for longevity than does the feet-legs composite.

Frequency of abortions recorded through Dairy Herd Improvement (DHI) testing was summarized for cows with lactations completed from 2001 through 2009. For 8.5 million DHI lactations of cows that had recorded breeding dates and were >151 d pregnant at lactation termination, the frequency of recorded abortions was 1.31%. Effects of year, herd-year, month, and pregnancy stage at lactation termination; parity; breed; milk yield; herd size; geographic region; and state within region associated with DHI-recorded abortion were examined. Abortions recorded through DHI (minimum gestation of 152 d required) were more frequent during early gestation; least squares means (LSM) were 4.38, 3.27, 1.19, and 0.59% for 152 to 175, 176 to 200, 201 to 225, and 226 to 250 d pregnant, respectively. Frequency of DHI-recorded abortions was 1.40% for parity 1 and 1.01% for parity ≥ 8. Abortion frequency was highest from May through August (1.42 to 1.53%) and lowest from October through February (1.09 to 1.21%). Frequency of DHI-recorded abortions was higher for Holsteins (1.32%) than for Jerseys (1.10%) and other breeds (1.27%). Little relationship was found between DHI-recorded abortions and herd size. Abortion frequencies for effects should be considered to be underestimated because many abortions, especially those caused by genetic recessives, go undetected. Therefore, various nonreturn rates (NRR; 60, 80, …, 200 d) were calculated to document pregnancy loss confirmed by the absence of homozygotes in the population. Breeding records for April 2011 US Department of Agriculture sire conception rate evaluations were analyzed with the model used for official evaluations with the addition of an interaction between carrier status of the service sire (embryo's sire) and cow sire (embryo's maternal grandsire). Over 13 million matings were examined using various NRR for Holstein lethal recessive traits (brachyspina and complex vertebral malformation) and undesirable recessive haplotypes (HH1, HH2, and HH3) as well as >61,000 matings for a Brown Swiss haplotype (BH1), and 670,000 matings for a Jersey haplotype (JH1). Over 80% of fertility loss occurred by 60 d after breeding for BH1, HH3, and JH1, by 80 d for HH2, by 100 d for BY, and by 180 d for HH1. For complex vertebral malformation, fertility loss increased from 40 to 74% across gestation. Association of undesirable recessives with DHI-recorded abortions ranged from 0.0% for Jerseys to 2.4% for Holsteins.

Noncompliance with current US and European Union (EU) standards for bulk-tank somatic cell count (BTSCC) as well as BTSCC standards recently proposed by 3 US organizations was evaluated using US Dairy Herd Improvement Association (DHI) herds and herds supplying milk to 4 Federal Milk Marketing Orders (FMO). Herds with 15 to 26 tests (frequently monthly) from January 2009 through October 2010 were included. Somatic cell scores (SCS) from 14,854 herds and 164,794 herd-tests were analyzed for DHI herds with ≥10 cows for all tests. Herd test-day SCC was derived as a proxy for BTSCC and was the basis for determining noncompliance and percentage of the milk it represented. For FMO herds, actual milk marketed and BTSCC were available from 27,759 herds and 325,690 herd-tests. A herd was noncompliant for the current EU BTSCC standard after 4 consecutive rolling 3-test geometric means (geometric method) were >400,000 cells/mL. A herd was noncompliant for the current US BTSCC standard after 3 of 5 consecutive monthly BTSCC shipments (frequency method) were >750,000 cells/mL. Alternative proposed standards (600,000, 500,000, or 400,000 cells/mL) also were examined. A third method designated noncompliance when a single 3-mo geometric mean of >550,000 or >400,000 cells/mL and a subsequent test exceeded the same level. Results were examined based on herd size or milk shipped by month. Noncompliance for the current US standard for the 12 mo ending October 2010 in DHI and FMO herds was 0.9 and 1.0%, respectively, compared with 7.8 and 16.1% for the current EU standard. Noncompliance was always greater for the frequency method than for the geometric method and was inversely related to herd size or milk shipped. Using the frequency method at 400,000 cells/mL, noncompliance was 19.1% for DHI herd-tests in herds with <50 cows compared with 1.1% for herds with ≥ 1,000 cows. For FMO herds shipping <900 t, noncompliance was 44.5% using the frequency method at 400,000 cells/mL compared with 8.0% for herds marketing >9,000 t. All methods proposed increased the percentages of herds and shipped milk that exceeded the regulatory limit. Producers will need to place more emphasis on reducing the incidence and prevalence of subclinical mastitis through known management practices such as proper milking techniques, well-functioning milking machines, postmilking teat disinfectant, dry cow treatment, and culling of problem cows to meet any of the proposed new standards.

Genetic evaluations for gestation length (GL) for Holstein service sires were studied to determine their effectiveness in predicting GL in an independent data set. Consequences of selection on GL were also assessed by examining correlated changes in milk and fitness traits. Holstein bulls with ≥ 300 calvings between 1998 and 2005 were stratified into the following 7 groups using predicted transmitting ability (PTA) for service sire GL: <-3.00, -3.00 to -2.01, …, 1.00 to 1.99, and ≥ 2.00 d. An independent set of 261,598 first-parity cows mated later to the same bulls and calving between 2006 and 2009 were segregated by the service sire PTA GL groups (group had 8,317 to 73,324 gestations), and these mates' GL were examined to determine effectiveness of service sire PTA GL. The model included fixed effects for herd-year and service sire group, plus covariates for conception dates to account for time opportunity among mates. Mean GL for mates by service sire group (from lowest to highest PTA GL) were 275.3, 276.5, 277.8, 278.6, 279.5, 280.6, and 281.7 d. Thus, service sire PTA GL was effective in identifying bulls that modified GL. Subsequent yield and fitness traits were also examined for the (independent) mates with the same service sire groups. Intermediate service sire PTA GL was optimal for yield traits and days open; performance for productive life and culling generally became less favorable as service sire PTA GL increased. A second examination was made by replacing service sire PTA GL groups in the model with phenotypic cow GL groups. Relationships between GL and subsequent performance for milk yield and fitness traits were examined using 9 phenotypic cow GL groups: ≤ 271, 272-273, …, 284-285, and ≥ 286 d. Performance generally improved for subsequent lactation yield as cow GL increased; however, intermediate GL was optimal for productive life, calving ease, stillbirth, culling, and days open. Results indicated that neither shortening nor increasing the mean for GL in the Holstein breed provided much overall benefit when all traits were considered. The same traits examined in the cows for the correlated effect from various GL were also examined in their offspring to determine whether the GL producing the calf had any influence on these same traits when the offspring reached their own productive period. Little carryover occurred from GL on the dam to the other traits observed on the offspring when examined a generation later.

Four US genetic-economic indices for dairy cattle were retrofitted to illustrate differences in phenotypic response observed for retrospective selection over 2 generations for currently evaluated traits, even though producers did not have evaluations available at the time for direct selection for those traits. Differences among cows were compared based on ranking of their sires and maternal grandsires (MGS) for the 4 retrofitted indices. Holstein artificial insemination bulls (106,471) were categorized by quintile for each index, and 25 cow groups were formed based on quintiles for sire and MGS (2 generations). Data included records from 1,756,805 cows in 26,106 herds for yield traits, productive life, pregnancy rate, and somatic cell score; 692,656 cows in 9,967 herds for calving difficulty; and 270,564 cows in 4,534 herds for stillbirths. For each index, least squares differences between the 25 cow groups were examined for 8 first-parity traits (milk, fat, and protein yields; productive life; somatic cell score; pregnancy rate; calving difficulty; and stillbirth) that had been standardized for age. Analysis removed effects of herd and cow birth year. Seven of 25 cow groups were consolidated into 3 groups based on index ranking for their male ancestors (low, medium, and high). The cow group with high sire and MGS rankings for the 2006 net merit index produced more milk (219 kg), fat (21 kg), and protein (11 kg) and had longer productive life (6.3 mo), lower somatic cell score (0.21), higher pregnancy rate (1.2 percentage units), fewer difficult births in heifers (3.8 percentage units), and lower stillbirth rate (4.6 percentage units) than did the group with low sire and MGS rankings. For cow groups based on sire and MGS rankings for 1971 (milk and fat) and 1977 (milk, fat, and protein) indices, advantages for the group with high sire and MGS rankings were much larger for yield traits but smaller (and sometimes even unfavorable) for other traits. Cow groups based on sire and MGS rankings for the 1994 net merit index generally had differences that were intermediate to groups based on sire and MGS rankings for the 1977 and 2006 indices. Phenotypic differences revealed retrospectively between genetic-economic indices indicate that genetic improvement should be made for all traits included in recent net merit indices.

Reproductive information since 1995 from the USDA national dairy database was used to calculate yearly Holstein and Jersey means for days to first breeding after calving (DFB), 70-d nonreturn rate, conception rate (CR), number of breedings per lactation (NB), interval between first and last breedings during the lactation, days to last breeding after calving (DLB), pregnancy rate (PR), calving interval (CI), and interval between consecutive breedings. Data were from nearly 20 million breedings during >8 million lactations of >5 million cows in >23,000 herds. Means were also calculated for some traits by parity and breeding number for both breeds and by geographical region and synchronization status for Holsteins. The DFB declined for Holsteins from 92 d in 1996 to 85 d in 2007; the trend in yearly differences was not as consistent for Jerseys. First- and all-breeding 70-d nonreturn rate declined 5 to 9 percentage units over time. First- and all-breeding CR declined 2 to 4 percentage units. The DFB were longer for later parities of Holsteins than for early parities. Second-and third-breeding CR were sometimes 1 to 2 percentage units above first-breeding CR for Holsteins but lower (1 to 7 percentage units) for Jerseys. The CR within breeding number declined across parities for both breeds. The NB increased by 0.3 to 0.4 breedings over time but remained constant (2.5 or 2.6 breedings) across parities for Holsteins and increased (from 2.2 to 2.4 breedings) for Jerseys. Holstein DFB were fewest in the Northwest (78 d) and greatest in the Mountain region (92 d). Regional CR was highest for the Northeast and Southwest (33%) and lowest for the Southeast (26%); NB was fewest for the Northeast (2.3) and greatest for the Southeast (2.7). Mean DLB was fewest for the Southwest (127 d) and greatest for the Mountain region (157 d); CI was shortest for the Southwest (406 d) and longest for the Mideast (434 d). Mean PR was highest for the Southwest (28.3%) and lowest for the Mideast and Southeast (22.2%). Use of timed artificial insemination following synchronizedestrus appears to have reduced DFB, lowered CR, and increased NB while reducing DLB and CI. However, synchronized breeding was not a primary cause of Holstein regional differences for reproductive traits. Since 2002, phenotypic performance for CR, DLB, and CI as well as genetic merit for daughter PR have stopped their historical declines and started to improve.

A retrospective study of the impact of the estimated breeding values of sires and maternal grandsires for somatic cell score (SCS) on productive life (PL) of Holsteins and Jerseys was conducted. Data included records from 2,626,425 Holstein and 142,725 Jersey cows. The sires and maternal grandsires of cows were required to have been available through artificial insemination and to have predicted transmitting ability (PTA) SCS evaluations based on 35 or more daughters. A weighted function (WPTA) of sire and maternal grandsire PTA for SCS was used: (sire PTA + 0.5 maternal grandsire PTA)/1.5. The 3 dependent variables were PL, frequency of cows culled for mastitis, and first-lactation SCS. The model included effects of herd, birth year, and WPTA (WPTA was categorized into groups: <2.70, 2.70 to 2.79, ..., 3.20 to 3.29, > or =3.30). For analysis of first-lactation SCS, calving year and calving month were substituted for birth year. Differences among WPTA groups were highly significant: as WPTA increased, PL decreased, whereas percentage culled for mastitis and first-lactation SCS increased. The range in PL from lowest to highest WPTA was 5.07 mo for Holsteins and 4.73 mo for Jerseys. Corresponding differences for percentage culled for mastitis were 7.0 and 5.6% and for SCS were 0.95 and 1.04 (for Holsteins and Jerseys, respectively). Although phenotypic studies suggest that cows with extremely low SCS were less resistant to mastitis, our results showed consistent improvements in PL, percentage culled for mastitis, and SCS of daughters when bulls were chosen for low PTA SCS.

Genetic and environmental factors that might affect gestation length (GL) were investigated. Data included information from >11 million parturitions from 1999 through 2006 for 7 US dairy breeds. Effects examined were year, herd-year, month, and age within parity of conception; parturition code (sex and multiple-birth status); lactation length and standardized milk yield of cow; service sire; cow sire; and cow. All effects were fixed except for service sire, cow sire, and cow. Mean GL for heifers and cows, respectively, were 277.8 and 279.4 d for Holsteins, 278.4 and 280.0 d for Jerseys, 279.3 and 281.1 d for Milking Shorthorns, 281.6 and 281.7 d for Ayrshires, 284.8 and 285.7 d for Guernseys, and 287.2 and 287.5 d for Brown Swiss. Estimated standard deviations of GL were greatly affected by data restrictions but generally were approximately 5 to 6 d. Year effects on GL were extremely small, but month effects were moderate. For Holstein cows, GL was 2.0 d shorter for October conceptions than for January and February conceptions; 4.7 and 5.6 d shorter for multiple births of the same sex than for single-birth females and males, respectively; 0.8 d longer for lactations of < or =250 d than for lactations of > or =501 d; and 0.6 d shorter for standardized yield of < or =8,000 kg than for yield of > or =14,001 kg. Estimates for GL heritability from parities 2 to 5 were 33 to 36% for service sire and 7 to 12% for cow sire; corresponding estimates from parity 1 were 46 to 47% and 10 to 12%. Estimates of genetic correlations between effects of service sire and cow sire on GL were 0.70 to 0.85 for Brown Swiss, Holsteins, and Jerseys, which indicates that those traits likely are controlled by many of the same genes and can be used to evaluate each other. More accurate prediction of calving dates can help dairy producers to meet management requirements of pregnant animals and to administer better health care during high-risk phases of animals' lives. However, intentional selection for either shorter or longer GL is not recommended without consideration of its possible effect on other dependent traits (e.g., calving ease and stillbirth).

In August 2007, the USDA changed from calculating official genetic evaluations quarterly to triannually in conjunction with the schedule change for international evaluations. To offset part of the delay in providing genetic information because of the reduced frequency of official evaluations, industry cooperators requested that interim evaluations be initiated for progeny-test (PT) bulls based on first-lactation records from PT daughters and their contemporaries that calved recently in cooperator herds. Alternatives for interim evaluations were studied to determine which would characterize genetic merit of PT bulls most accurately. Four alternative Holstein data sources were examined based on maximum data interval (most recent 12 or 18 mo of first calvings) and minimum number of PT daughters in herd (> or =1 or > or = 5). The highest correlation between August 2006 interim and official evaluations for milk yield was 0.980 for interim evaluations based on the most recent 18 mo of first calvings from cooperator herds with > or =1 PT daughter. That high correlation confirmed that interim evaluations based on limited data could provide genetic estimates of value between official evaluations. With the support of the Council on Dairy Cattle Breeding, the USDA initiated 3 interim evaluations each year with release limited to PT bulls with > or =10 daughters and an increase in reliability since the most recent official evaluation.

Factors that affect frequency of death of lactating cows were studied for cows with records that terminated from 1995 through 2005. Analyses included effects of herd, year, month, parity, and lactation stage at lactation termination as well as cow breed and milk yield. A national data set (15,025,035 lactations in 45,032 herds) was analyzed with PROC GLM. Overall death frequency was 3.1% per lactation (5.7% per cow). Death frequency increased by 1.6% from 1995 to 2005, with a sudden increase of 0.9% from 2003 to 2004, probably because of a USDA requirement in late 2003 for euthanizing downer cows. Death frequency was 16.5% greater for lactations that terminated at or=251 d. Death frequency increased with parity (2% greater for eighth parity and later than for first parity) and with lactation milk yield (0.4%/1,000 kg for Holsteins and Jerseys and 0.5%/1,000 kg for other breeds). Deaths were most frequent in July and least frequent in November. Within-herd breed differences (Holstein, Jersey, and other breeds) were small. The heritability of likelihood of death estimated from a sample of 79,162 Holstein cows was 1.3%. Death losses are increasing, perhaps partly because of increased milk yield and more intensive management regimens.

Objective was to improve the accuracy of computer programs that optimize breeding and replacement decisions for individual cows to support economic decision making. Most recently published optimization programs characterize cows by a limited number of discrete states, a combination of (at least) parity, level of milk production, stage of lactation, and pregnancy status. The number of states is necessarily limited to make computations feasible. Consequently, cows may not be well represented by a state and the program’s decisions may be significantly different from optimal. For example, a small number of lactation curves (for example Wood’s curves) are typically used depending on parity and level of milk production. In practice, every cow has a unique lactation curve. A significant improvement in accuracy of prediction of cow performance can be obtained when the states for the current lactations of the cows currently in the herd are separated from the states for their next lactation and replacement heifers. The states for cows currently in the herd can then be dramatically increased with few extra computations. These adjustments allow for much more accurate predictions of individual cow performance in the current lactation than previously available. Predicted cow performance for the remainder of the lactation can be updated daily. Consequently, optimized breeding and culling decisions are more accurate. To illustrate, 400 cows from the University of Florida dairy research herd were used. Individual lactation curves were predicted by Best Prediction. The resulting 400 lactation curves predicted milk yields in the remainder of the lactation better than the few standard Wood’s curves in traditional programs. Optimal breeding and culling decisions were frequently different. The program currently runs as an add-on to existing dairy herd information systems using automated data exchange. The proposed improvement in accuracy is an important step towards more realistic computerized decision support on dairy farms.