Screening of key traits for indirect selection and prediction model construction of intramuscular fat content based on machine learning in live pigs
[Objectives]The aim of this study was to investigate the effects of weaning weight,live weight before slaughter,backfat thickness and other traits on intramuscular fat(IMF)content in pigs,and to determine the key trait factors influencing IMF content in pigs.[Methods]A total of 805 Pietrain ×[Duroc ×(Landrace × Large White)]four-cross commercial pigs were used in this study,and the sex was recorded,and 14 traits including birth weight,weaning weight,live weight before slaughter,and IMF content were measured.First,the trait factors affecting IMF content were preliminarily screened out from 14 traits by correlation analysis.Then,the random forest model was used to evaluate the significance of each trait factor on IMF content,and the key trait factors affecting IMF were further screened by LASSO regression and stepwise regression.On this basis,the generalized linear model(GLM)was used to analyze the effects of different levels of key trait factors on IMF content.[Results]Correlation analysis showed that IMF content was significantly correlated with weaning weight(r=0.13,P＜0.001)and live weight before slaughter(r=0.22,P＜0.001).Backfat thickness at different locations was significantly correlated with IMF content(P＜0.001)with correlation coefficients ranging from 0.21 to 0.26(P＜0.001).In addition,the IMF content was also significantly correlated with the red value(a*),yellow value(b*),hue angle(H0)and chroma(C*)of the meat color,and the correlation coefficients were ranged from 0.08 to 0.13(P＜0.05).Random forest model analyses showed that backfat thickness at the thoracolumbar junction contributed the most to IMF content,followed by pre-slaughter live weight.LASSO regression and stepwise regression were used to screen nine and five trait factors being identified as significantly affecting IMF content,respectively,and the four in vivo measurable traits,namely,sex,weaning weight,pre-slaughter live weight,and thoracolumbar conjunctive backfat thickness,were the key trait factors screened by the two methods in total.GLM analysis showed that all four live measurable traits had a significant effect on IMF content,and the average IMF content of castrated gilts(2.52％)was significantly higher than that of sows(2.41％)(P＜0.05);the average IMF content of the group with weaning weight less than 5 kg(2.24％)was significantly lower than those of the other three groups(P＜0.05);and the average IMF content of the group with live weight before slaughter less than 85 kg(2.27％)was significantly lower than that of the group with live weight above 115 kg(2.67％)(P＜0.05).IMF content(2.27％)was significantly lower(P＜0.05)than that of the group with a pre-slaughter live weight of more than 115 kg(2.67％),and when the pre-slaughter live weight was greater than 100 kg,the difference in average IMF content between the level groups was not significant(P＞0.05).In addition,the average IMF content of the group with a backfat thickness greater than 26 mm at the thoracolumbar vertebral junction(2.73％)was significantly higher than those of the other backfat thickness groups(P＜0.05),whereas the difference between the average IMF content of the groups with a backfat thickness of 5-12 mm and 12-19 mm was not significant(P＞0.05).[Conclusions]In this study,four in vivo measurable traits,namely sex,weaning weight,pre-slaughter live weight and backfat thickness at the thoracolumbar vertebral junction,which were significantly correlated with IMF content,were identified by machine learning,and a significant upward trend of average IMF content with increasing pre-slaughter live weight and backfat thickness at the thoracolumbar vertebral junction was found.
万方数据
维普
