Search

Dairy farmers are motivated to ensure cows become pregnant in an optimal and timely manner. Although timed artificial insemination (TAI) is a successful management tool in dairy cattle, it masks an animal's innate fertility performance, likely reducing the accuracy of genetic evaluations for fertility traits. Therefore, separating fertility traits based on the recorded management technique involved in the breeding process or adding the breeding protocol as an effect to the model can be viable approaches to address the potential bias caused by such management decisions. Nevertheless, there is a lack of specificity and uniformity in the recording of breeding protocol descriptions by dairy farmers. Therefore, this study investigated the use of 8 supervised machine learning algorithms to classify 1,835 unique breeding protocol descriptions from 981 herds into the following 2 classes: TAI or other than TAI. Our results showed that models that used a stacking classifier algorithm had the highest Matthews correlation coefficient (0.94 ± 0.04, mean ± SD) and maximized precision and recall (F1-score = 0.96 ± 0.03) on test data. Nonetheless, their F1-scores on test data were not different from 5 out of the other 7 algorithms considered. Altogether, results presented herein suggest machine learning algorithms can be used to produce robust models that correctly identify TAI protocols from dairy cattle breeding records, thus opening the opportunity for unbiased genetic evaluation of animals based on their natural fertility.

International audience; To optimize breeding for feed efficiency, dry matter intake data from 3,967 Holstein cows with 131,234 weekly records from Denmark, France, The Netherlands, and Canada were combined. A random regression model was fitted to 1 st and 2 nd parity data for dry matter intake, energy corrected milk and body weight. Results showed that genetic residual feed intake is heritable in both parities, with an average heritability of 0.21. Genetic correlations across lactation and parities showed moderate to high estimates between mid and late lactation within parity. Early lactation had low genetic correlation with mid and late lactation. The genetic correlations between parities were around zero at all lactation stages. These findings show the importance of data collection and modelling at all lactation stages and parities.

Oral Presentation at the 2020 ADSA Annual Meeting (online). The objectives of this study were to identify the temperature-humidity index (THI) thresholds where heat load starts to affect production traits (milk, fat, and protein yield) in the first 3 lactations of Holstein cows, and to assess potential differences in heat stress in Quebec and Ontario. A total of 2.1 million test-day records from 167,620 Ontario cows and 3.1 million records from 225,104 Quebec cows spanning an 11-yr period (2008 – 2018) were used. The climatic data consisted of hourly measurements of ambient temperature and relative humidity. In total 58 weather stations were located within a maximum distance of 20km from each herd. A k-means cluster analysis was performed within each THI >75 to exclude herds that may have some type of cooling system. A linear model was fitted to adjust the phenotypes, which were then plotted against the THI levels for each parity. To identify the THI threshold at which milk production starts to decline, a segmented polynomial was used to describe the shape of the curve of the relationship between production and THI values. During the summer, the average THI was 69 and 68 in Ontario and Quebec, respectively. Two heat stress thresholds were identified for milk yield and protein yield, one being associated with a low decline in the yield and the other one associated with sharper decline. For both provinces, the second threshold was reached at about THI = 79 for both milk and protein yields, while the first threshold was reached at about a THI = 64 – 72 for milk yield and THI = 57 – 64 for protein yield, depending on the province and parity. The expected reduction in milk yield after a THI ~79 was −0.40 kg/d and −0.19kg/d in Ontario and Quebec, respectively. The same features for protein yield were −16.2 g/d and −9.7 g/d. For fat yield, the expected reduction in yield after a THI ~57 was −3.3g/d and −2.6g/d for Ontario and Quebec, respectively. Therefore, the estimated rate decline differed between the 2 provinces, which may be related to the higher percentage of tie stall barns in Quebec (92%) compared with Ontario (

Oral Presentation at the 2020 ADSA Annual Meeting (online). Breeding programs across the globe are shifting their selection goals from production-driven to more balanced indexes that include traits of direct and indirect economic significance, such as conformation, fertility and health. In Canada, for instance, changes were made to better reflect producers and market needs in the wake of improved accuracy of genetic and genomic evaluations and national efforts regarding data recordings. In 2015 a new national profit index, Pro$, was introduced to maximize production yields, while maintaining functional and conformation traits, taking into consideration the accumulated profit to 6 years of age of proven sire’s daughters. The Canadian classification system comprises several linear and nonlinear descriptive traits and defective characteristics that are taken into account to calculate points, which are distributed in 4 different scorecard sections: Mammary System (MS), Feet & Legs, Dairy Strength (DS) and Rump. Understanding the contribution of each of these traits to a monetary index, such as Pro$, would help producers make culling and mating decisions ultimately resulting in more profitable herds. Therefore, the present study used statistical regression and principal component analysis to look at the overall and individual contribution of 23 classification traits to the Pro$ index using deregressed Estimated Breeding Values from 9,292 proven bulls (Lactanet). The total variance explained by the full model was 0.66, and out of the traits included in the model, Dairy Capacity contributed the most to the increase in Pro$. Stepwise backward regression analysis indicated that all traits should remain in the full model. Principal component analysis showed that the vast majority of traits contributed positively to Pro$ values. Traits from the MS contributed the most to the increase in Pro$ values, whereas Thurl Placement and Rump Angle did not seem to have a sizeable impact on Pro$. Dairy Research Cluster 3 – Canada, Canadian Agricultural Partnership, Lactanet, Center for Genetic Improvement of Livestock (CGIL)

International audience