Estimates of the effect of pregnancy on production traits of Canadian dairy breeds

Estimates of the effect of pregnancy on production traits of Canadian dairy breeds J. Bohmanova*, F. Miglior†‡, J. Jamrozik* * Centre for Genetic Imp...
0 downloads 2 Views 710KB Size
Estimates of the effect of pregnancy on production traits of Canadian dairy breeds J. Bohmanova*, F. Miglior†‡, J. Jamrozik* *

Centre for Genetic Improvement of Livestock, Department of Animal and Poultry Science, University of Guelph, Guelph, Ontario, Canada N1G 2W, Canada † Agriculture and Agri-Food Canada, Sherbrooke, QC, Canada ‡ Canadian Dairy Network, Guelph, ON, Canada

Introduction Pregnancy has been reported to have a negative effect on milk yield of dairy cows due to hormonal changes, causing regression of the mammary gland (Akers, 2006), and nutrient requirements of the fetus, reducing available nutrients for milk production (Bell et al., 1995). The effect of pregnancy is small at the beginning of gestation and becomes greater at later stages of gestation, when growth and nutrient requirements of the conceptus are larger. Not accounting for the effect of pregnancy can lead to overestimation of breeding values of non-pregnant cows. Bohmanova et al. (2008) compared seven models with different ways of accounting for the effect of pregnancy. The most suitable model for accounting for the effect of pregnancy on production of Canadian Holstein cows was a model including an effect of stage of pregnancy. Recommendations were to estimate the effect of pregnancy from records where conception date was confirmed by AI records and subsequent calving; and then to use these estimates to preadjust test-day records that are used for genetic evaluation. The objectives of this study were: 1) to estimate the effect of pregnancy on production traits and 2) to develop additive adjustment factors for adjusting milk, fat and protein test-day yields from first three lactations of Ayrshire, Brown Swiss, Canadienne, Guernsey, Holstein, Jersey and Milking Shorthorn cows for the effect of pregnancy. Data Data were 52,147,841 test-day (TD) milk, fat and protein yield and somatic cell score records from the first three lactations of Ayrshire, Brown Swiss, Canadienne, Guernsey, Holstein, Jersey and Milking Shorthorn cows calved from 1998 to 2007. Only TD records from 5 to 365 DIM were used. Description of the data is given in Table 1. Cows were classified according to insemination and subsequent calving data availability into 5 classes (A1-A5). A gestation length (GL) of 280-d was assumed for Ayrshire, Canadienne, Holstein, Jersey and Milking Shorthorn, and 285-d for Brown Swiss and Guernsey breed. When a cow had a subsequent lactation but no breeding record, the conception date was set to the date 280-d for Ayrshire, Canadienne, Holstein, Jersey and Milking Shorthorn, and 285-d for Brown Swiss and Guernsey breed prior to her subsequent calving (A2). For a cow with subsequent calving and insemination records located in the interval of GL±15-d prior to her subsequent calving (A1), the conception date was set to the date of her last insemination record in this interval. For a cow without subsequent calving but with available insemination records after her last calving, the conception date was set to her last available insemination record (A3). A cow with completed lactation but without a subsequent calving and without insemination records was assumed to be non-pregnant (A4). Last available TD record was assumed to be the first day of   1

dry period for a cow with lactation in progress (A5). Considering an average dry period of 60 days, the conception date was set to (GL-60) days prior the last TD record. If such conception date occurred earlier than 125 DIM, it was set to 125-d after calving date. Cows with days open shorter than 14-d and longer than 500-d were excluded. Test-day records were divided into 13 stages of pregnancy classes defined as W1 (days pregnant ≤ 10), W2 (11≤ days pregnant ≤31), W3 (32≤ days pregnant ≤52), W4 (53≤ days pregnant ≤73), W5 (74≤ days pregnant ≤94), W6 (95≤ days pregnant ≤115), W7 (116≤ days pregnant ≤136), W8 (137≤ days pregnant ≤157), W9 (158≤ days pregnant ≤178), W10 (179≤ days pregnant ≤199), W11 (200≤ days pregnant ≤220), W12 (221≤ days pregnant ≤241), W13 (days pregnant≥242). Table 1: Description of data Number of Breed test-day records Ayrshire 1,915,261 Brown Swiss 229,923 Canadienne 43,979 Guernsey 155,300 Holstein 48,638,184 Jersey 1,137,647 Milking Shorthorn 27,547

Number of cows 105,725 13,503 2,660 9,347 2,826,456 67,276 1,826

Average daily milk yield (kg) Lactation 1 Lactation 2 Lactation 3 19.5 22.0 23.2 20.2 23.5 24.9 14.3 16.6 17.8 17.9 19.9 20.4 25.4 29.3 30.9 16.8 19.4 20.6 18.4 21.2 22.6

Materials and Methods The model used for estimation of the effect of pregnancy was a multiple-trait, multi-lactation random regression model defined as (Bohmanova at al., 2008): 6

4

4

yijklmpw = HTDij + ∑ α ikn zn (dim) + dim_cll + ∑ βimn zn (dim) + ∑ γ imn zn (dim) + υ pw + eijklmpw , n=0

n=0

n =0

6

4

4

n=0

n =0

n=0

 

y *ijklmpw = HTDij + ∑ α ikn zn (dim) + dim_cll + ∑ β imn zn (dim) + ∑ γ imn zn (dim) + eijklmpw ,

where y was either milk, fat or protein yield and y* was SCS, HTDij was the jth herd-test-date fixed effect for a trait i (TD milk, fat, protein yield and SCS), αikn was the nth fixed regression coefficient for the ith trait specific to the nth region-age-season class, dim_cll was the fixed effect of the lth DIM class (l=5,…..365), βirn was the nth random regression coefficient for the additive genetic effect of animal m, γirn was the nth random regression coefficient for the permanent environmental effect of cow m, z(dim) was the vector of fixed and random regressions evaluated at DIM dim. Legendre polynomials of order 6 and 4 were used to fit fixed and random regressions, respectively; υw was the effect of the pth source of information class (A1-A5) and the wth stage of pregnancy class. The effect of pregnancy was not included in the model for SCS. Variance components used in this study were estimated in a separate study by a Gibbs Sampler using subsets of approximately 100,000 test-day records for each breed, except of Canadianne and Milking Shorthorn, where all available records were used. Miglior et al. (2008) showed that the effect of pregnancy does not have a significant effect on variance components. The model

  2

used for estimation of variance components was therefore a simplified model where the effect of pregnancy was not considered. Mixed model equations were solved by iteration on data with a Preconditioned Conjugate Gradient algorithm using a block diagonal preconditioner (Lidauer et al., 1999). As shown in Table 1, the seven Canadian dairy breeds have different average milk yield. Direct comparison of the effect of pregnancy between breeds would not be therefore a correct comparison. A proportional effect of pregnancy (PEP) on total milk yield was introduced in order to address differences in milk yield among breeds. It was defined as a ratio of milk yield lost per pregnancy and the total milk yield. The total milk yield was calculated as an average milk yield multiplied by 351. The number 351 was based on an assumption that an average cow is 130 days open, has a gestation length of 285-d, days dry of 60-d, and a lactation length of 355d= (130+(285-60)) and because milk (colostrums) is not collected until the 5th day in milk, such a cow would produce milk for 351 days. The formula for calculating PEP was: (285 − 60)

PEPik =

∑ j =1

225

pregijk

avgM ik × (130 + (285 − 60) − 4)

×100 =

∑ preg j =1

ijk

avgM ik × 351

×100,

where pregijk is the effect of pregnancy for the ith lactation, kth breed and jth days pregnancy and avgMikik is the average daily milk yield of the kth breed in ith lactation (Table 1). Adjustment factors were calculated by interpolating estimates of the effect of pregnancy obtained from A1 class (conception date calculated from both AI and subsequent calving records) as: v −v pregt = vi + j i × (t − mi ) for mi ≤ t < m j , m j − mi where pregt is the adjustment factor for the tth day of pregnancy, mi and mj are midpoints of the stage of pregnancy classes surrounding the point t and νi and νj are the estimates of the effect of pregnancy for these classes (Figure 1). Results The relative effect of pregnancy, expressed as PEP, was the highest in the Ayrshire and lowest in the Jersey breed (Table 2). The size of the effect increased with the lactation number. In Holstein, 2.6%, 2.6% and 3.2% of milk yield was lost due to pregnancy in first, second and third lactation, respectively. Table 2: Proportional effect of pregnancy (PEP) on milk yield of Ayrshire, Brown Swiss, Guernsey, Holstein and Jersey Breed Lactation 1 Lactation 2 Lactation 3 Ayrshire 3.2% 3.3% 3.8% Brown Swiss 2.9% 3.3% 2.6% Guernsey 2.9% 2.6% 3.6% Holstein 2.6% 2.6% 3.2% Jersey 1.9% 2.0% 2.2% As shown in Figures 2, 3, 6 and 7, milk yield of Ayrshire, Brown Swiss, Guernsey and Holstein cows declined slowly in the first 105-d of pregnancy and then the slope of the decline   3

increased. In Canadienne and Milking Shorthorn (Figures 4 and 9), the milk yield slightly increased in the first four months. This doesn’t have a biological explanation and it was most likely caused by a limited number of records. The estimates of Canadienne and Milking Shorthorn were therefore corrected by assuming a null effect of pregnancy in the first 4 months. The effect of pregnancy in the later stages was created by fitting a cubic regression through the estimates from the model (Figure 5 and 10).The same corrections were applied to the estimates for fat and protein of these two breeds (Figures 13, 17, 20 and 24). In Jerseys (Figure 8), milk yield also increased in the first three months of pregnancy but because the increase was very small (

Suggest Documents