Cluster analysis and potential influencing factors of boars with different fertility.

The fertility of boars is intimately tied to the pig farm's economic benefits. This study aimed to rapidly categorize boars of different fertility and investigate the factors influencing the categorization using the production data in a large pig farm in northern China, including 11,163 semen collection records of Yorkshire boars (215), 11,163 breeding records and 8770 records of farrowing performance of Yorkshire sows (4505), as well as 4720 records of selection indices (sire line index and dam line index) for boars and sows (215 and 4505) between 2017 and 2020. The boar population was classified by two-step cluster analysis, followed by factor analysis to minimize the dimensionality of data variables and eliminate multicollinearity, and then using ordinal logistic regression model to investigate the risk variables impacting boar fertility categorization. Results showed that the two-step clustering divided the 215 boars into three subgroups: high-fertility (n = 61, 28.4%), medium-fertility (n = 127, 59.1%) and low-fertility (n = 27, 12.6%). The high-fertility boars were shown to be substantially greater than the medium-fertility or low-fertility boars (p < 0.05) in average total litter size, number of born alive, and number of healthy piglets of mated sows. Compared with low-fertility boars, the high-fertility boars were also significantly higher (p < 0.05) in the pregnancy rate and farrowing rate of mated sows. However, the three boar subgroups showed no difference (p > 0.05) in semen quality information (average sperm motility, average sperm density, and average sperm volume). Collinearity diagnosis indicated severe multicollinearity among the 20 data variables, which were reduced to 8 factor variables (factors 1-8) by factor analysis, and further collinearity diagnosis exhibited no multicollinearity among the 8 factor variables. Ordered logistic regression analysis revealed a significant and positive correlation (p < 0.05) of boar fertility with factor 2 (average total litter size, number of born alive, number of healthy piglets), factor 4 (average number of weak piglets and average weak piglet rate), factor 6 (sire line index of boars and dam line index of boars), factor 8 (pregnancy rate and farrowing rate), highlighting factor 2 as the most important factor influencing the classification of boar fertility. Our results indicate that the two-step cluster analysis can be used as a simple and effective method to screen boars with different fertility and that farm producers should pay attention to the recording of the reproductive performance of the mated sows due to its role as the risk factor for classification of boar fertility.