Genetic Analysis of Clinical Mastitis Data from On-Farm Management Software Using Threshold Models

Producer-recorded clinical mastitis data from 77,791 cows in 418 herds were used to determine the potential for genetic improvement of mastitis resistance using data from on-farm management software programs. The following threshold sire models were applied: 1) a single-trait lactation model, where mastitis was recorded as 0 or 1 in first lactation only; 2) a 3-trait lactation model, where mastitis was recorded as 0 or 1 in each of the first 3 lactations, and 3) a 12-trait, lactation-segment model, where mastitis was recorded as 0 or 1 in each of 4 segments (0 to 50, 51 to 155, 156 to 260, and 261 to 365 d postpartum) in each of the first 3 lactations. Lactation incidence rates were 0.16, 0.20, and 0.24 in first, second, and third lactation, respectively, and incidence rates within various segments of these lactations ranged from 0.036 in late first lactation to 0.093 in early third lactation. Estimated heritability of liability to clinical mastitis ranged from 0.07 to 0.15, depending on the model and stage of lactation. Heritability estimates were higher in first lactation than in subsequent lactations, but estimates were generally similar for different segments of the same lactation. Genetic correlations between lactations from the 3-trait model ranged from 0.42 to 0.49, while correlations between segments within lactation from the 12-trait model ranged from 0.26 to 0.64. Based on the results presented herein, it appears that at least 2 segments are needed per lactation, because mastitis in early lactation is lowly correlated with mastitis in mid or late lactation. Predicted transmitting abilities of sires ranged from 0.77 to 0.89 for probability of no mastitis during the first lactation and from 0.36 to 0.59 for probability of no mastitis during the first 3 lactations. Overall, this study shows that farmer-recorded clinical mastitis data can make a valuable contribution to genetic selection programs, but additional systems for gathering and storing this information must be developed, and more extensive data recording in progeny test herds should be encouraged.

Genetic selection for health traits using producer-recorded data. II. Genetic correlations, disease probabilities, and relationships with existing traits

The objectives of this study were to calculate genetic correlations between health traits that were recorded in on-farm herd management software programs and to assess relationships between these traits and other traits that are routinely evaluated in US dairy sires. Data consisted of 272,576 lactation incidence records for displaced abomasum (DA), ketosis (KET), mastitis (MAST), lameness (LAME), cystic ovaries (CYST), and metritis (MET) from 161,622 cows in 646 herds. These data were collected between January 1, 2001 and December 31, 2003 in herds using the Dairy Comp 305, DHI-Plus, or PCDART herd management software programs. Binary incidence data for all disorders were analyzed simultaneously using a multiple-trait threshold sire model that included random sire and herd-year-season of calving effects. Although data from multiple lactations were available for some animals, our genetic analysis included only first parity records due to concerns about selection bias and improper modeling of the covariance structure. Heritability estimates for the presence or absence of each disorder during first lactation were 0.14 for DA, 0.06 for KET, 0.09 for MAST, 0.03 for LAME, 0.04 for CYST, and 0.06 for MET. Estimated genetic correlations were 0.45 between DA and KET, 0.42 between KET and CYST, 0.20 between MAST and LAME, 0.19 between KET and LAME, 0.17 between DA and CYST, 0.17 between KET and LAME, 0.17 between KET and MET, and 0.16 between LAME and CYST. All other correlations were negligible. Correlations between predicted transmitting abilities for the aforementioned health traits and existing production, type, and fitness traits were low, though it must be noted that these estimates may have been biased by low reliability of the health trait evaluations. Based on results of this study, it appears that genetic selection for health disorders recorded in on-farm software programs can be effective. These traits can be incorporated into selection indices directly, or they can be combined into composite traits, such as "reproductive disorders", "metabolic disorders", or "early lactation disorders".

Genetic selection for health traits using producer-recorded data. I. Incidence rates, heritability estimates, and sire breeding values

The objective of this study was to determine the feasibility of genetic selection for health traits in dairy cattle using data recorded in on-farm herd management software programs. Data regarding displaced abomasum (DA), ketosis (KET), mastitis (MAST), lameness (LAME), cystic ovaries (CYST), and metritis (MET) were collected between January 1, 2001 and December 31, 2003 in herds using Dairy Comp 305, DHI-Plus, or PCDART herd management software programs. All herds in this study were either participants in the Alta Genetics (Watertown, WI) Advantage progeny testing program or customers of the Dairy Records Management Systems (Raleigh, NC) processing center. Minimum lactation incidence rates were applied to ensure adequate reporting of these disorders within individual herds. After editing, DA, KET, MAST, LAME, CYST, and MET data from 75,252 (313), 52,898 (250), 105,029 (429), 50,611 (212), 65,080 (340), and 97,318 (418) cows (herds) remained for analysis. Average lactation incidence rates were 0.03, 0.10, 0.20, 0.10, 0.08, and 0.21 for DA, KET, MAST, LAME, CYST, and MET (including retained placenta), respectively. Data for each disorder were analyzed separately using a threshold sire model that included a fixed parity effect and random sire and herd-year-season of calving effects; both first lactation and all lactation analyses were carried out. Heritability estimates from first lactation (all lactation) analyses were 0.18 (0.15) for DA, 0.11 (0.06) for KET, 0.10 (0.09) for MAST, 0.07 (0.06) for LAME, 0.08 (0.05) for CYST, and 0.08 (0.07) for MET. Corresponding heritability estimates for the pooled incidence rate of all diseases between calving and 50 d postpartum were 0.12 and 0.10 for the first and all lactation analyses, respectively. Mean differences in PTA for probability of disease between the 10 best and 10 worst sires were 0.034 for DA, 0.069 for KET, 0.130 for MAST, 0.054 for LAME, 0.039 for CYST, and 0.120 for MET. Based on the results of this study, it appears that genetic selection against common health disorders using data from on-farm recording systems is possible.

Genetic Evaluation of Dairy Sires for Milking Duration Using Electronically Recorded Milking Times of Their Daughters

The potential of using electronically recorded data from on-farm milking parlor and herd management software programs for genetic evaluation of dairy sires for milking duration of their daughters was assessed in the present study. Single measurements of milking duration were collected weekly from 29 herds between June 1, 2003 and April 1, 2004. These included 73,547 observations corresponding to 10,152 Holstein cows from 1551 sires. Average milking duration for a single milking in our data set was 4.5 min. Estimated heritability of milking duration was 0.17, and predicted transmitting abilities (PTA) of individual sires ranged from -0.48 min for sires with the fastest milking daughters to 0.59 min for sires with the slowest milking daughters. The correlation between PTA for milking duration and PTA for somatic cell score (SCS) was -0.15, indicating that sires whose daughters milk most quickly also tend to transmit higher SCS to their progeny. Correlations between PTA milking duration and PTA for teat placement and teat length were -0.14 and 0.20, respectively, indicating that sires that transmit wide teat placement and long teats tend to have daughters that milk slowly. Based on the results presented herein, it appears that genetic selection based on objective, electronically recorded milking times is possible. This approach would greatly improve the quality and efficiency of data collection relative to conventional evaluations of milking speed, which are based on farmer surveys. The number of herds currently equipped to routinely capture milking times is limited, but this number is increasing very rapidly. Future research should focus on refinement of data reporting and validation systems, as well as estimation of the economic value of milking duration. This trait may have an intermediate optimum, because cows that milk too slowly will disrupt parlor flow and reduce milking efficiency, but cows that milk too quickly may be at greater risk for mastitis.

Genetic parameters for yield traits of Holsteins, including lactose and somatic cell score

First lactation milk, fat, protein, and lactose yields and percentage yields were analyzed using a multiple-trait sire model including herd-year-season, sire group, and age of cow as fixed effects. Somatic cell score was fit both as a fixed effect in the model and as an additional dependent variable in two analyses with almost identical results. Variance components were estimated using REML with an expectation-maximization algorithm including sire relationships. Lactose means ranged from 4.84 to 4.97% across three dairy breeds. Data used to estimate variance components were first lactation Holstein records collected from 1986 to 1988 from 5246 daughters of 392 AI sires. Heritability estimates were .30, .29, .27, and .26 for milk, fat, protein, and lactose yields; .45, .47, and .53 for fat, protein, and lactose percentage yields; and .16 for somatic cell score. Genetic correlations of lactose percentage with milk, fat, protein, fat and protein percentages, and somatic cell score were -.30, -.16, -.21, .10, .29, and -.11, respectively, and phenotypic correlations were -.08, -.02, .01, .11, .29, and -.15.