Contact-free Measurements Of Conformation Traits And Genome-wide Association Analysis Of Body Size Traits In Sheep

Precision animal husbandry is an inevitable trend of animal husbandry modernization and a systematic innovation that runs through the whole breeding production system.As one of the most important livestock,sheep’s identity information and physical parameters are important indicators of precision breeding.In order to improve the traditional measurement methods,such as high cost,high stress response and other problems,this paper constructed the image data set of sheep body and face,combined with the image processing technology to measure the body and face features.Finally,with body weight as one of the covariables,genome-wide association analysis was conducted using the body size traits obtained by contact-free measurements,and multiple SNPs loci were found to be significantly correlated with body size traits.The specific results are as follows:(1)Based on side body images of 221 East Friesian sheep × Hu sheep(Donghu)F2generation lambs,contact-free measurements of body height,body length and chest depth were achieved after feature points were marked.Since the ratio between the length on the image and the real length is unknown,the three body scales measured in the image are converted into body length index,limb length index,limb length chest depth ratio and chest depth body length ratio.24 Donghu F2 generation lambs were randomly selected to compare the body measurement ratio measured in images with the actual body measurement ratio,and the mean absolute percentage error(MAPE)was 1.85% to 4.40%,indicating that the contact-free measurement method in this study has little error with the actual body size data.(2)Based on the body height,body length and chest depth data obtained from contact-free measurements,weight prediction models were established and the accuracy were assessed.Among 218 individuals of Donghu F2 generation with image-measured body height,body length,chest depth and actual weight data,198 groups of data were randomly selected as the training group for model establishment,and the remaining 20 groups of data were selected as the verification group.Four methods of simple regression,partial least squares,support vector regression and BP neural network regression were used to establish the body weight prediction model.Comparing the predicted results of each model with the actual weight measurement data,the results show that the weight prediction model based on support vector regression is the most accurate,with the absolute error ranging from-2.961 kg to +5.187 kg,and the average absolute percentage error is 9.5%.(3)The 4 body measurement ratio data of 167 Donghu F2 generation individuals were obtained by using the above contaction-free measurement method,and the whole genome of corresponding individuals was resequenced.Through GWAS analysis of body measurement ratio,26 SNPs were identified that were significantly correlated with limb length chest depth ratio,and 3 SNP sites that were significantly correlated with body length index.No significant correlation sites between limb length index and chest depth body length ratio were detected.(4)In addition,this study also attempted to label sheep facial feature points and quantitatively describe sheep facial features.The front and side image data sets of 20 Donghu F1 rams and 80 Hu ewes were constructed.After marking the feature points,alignment and normalization were carried out,and the combination of different feature points was used to quantitatively describe the front and side features of sheep.By further comparing the differences of facial features between individuals of different genders and breeds,it was found that the eye-lip Angle in positive features and the eye-nose Angle in lateral features had extremely significant differences between Donghu F1 rams and Hu ewes,suggesting that these two features could be used as references for the selection of subsequent facial recognition feature regions of sheep.In conclusion,the results of this study indicate that the image-based contact-free measurement method can obtain effective body size data and facial features,providing a new idea for the subsequent large-scale GWAS analysis of sheep body size traits and the localization of key genes.At the same time,it provides a theoretical basis for sheep facial recognition,and provides a scientific basis for the development of precision animal husbandry.