Abstract: The study aimed to compare the predictive accuracy of genomic selection for growth-related traits in sika deer(Cervus Nippon) based on models such as GBLUP. The heritability of growth-related traits such as body weight and body size of sika deer was estimated for 261 sika deer from a deer farm in Jilin Province from 2014 to 2019 (96 males and 165 females). Based on 5-fold cross-validation method, the prediction accuracy of GBLUP, Bayes A, Bayes B, Bayes C, Bayes Lasso and RRBLUP genomic selection models were compared. The suitable genome selection model for the growth-related traits of sika deer was screened. The results showed that:1) The heritability of shin circumference and high hips were 0.43 and 0.50, respectively, which belonged to high heritability; the heritability of body weight, body height and body length were 0.22, 0.30 and 0.27, respectively, which belonged to medium heritability; while the heritability of chest girth was 0.15, which belonged to low heritability. 2) In GBLUP, there was a positive correlation between the accuracy of genomic selection prediction and heritability of traits, but there was no significant positive correlation in Bayes and RRBLUP methods. 3) When the sample size was small, selecting GBLUP as the model of genome selection had certain advantages, and Bayes A could be used as the first choice for traits with low heritability. The highest prediction accuracy of body weight, body height, body length, shin circumference, chest girth and high hips were GBLUP, Bayes B, Bayes C, Bayes B, Bayes A and RRBLUP, respectively. In the actual production, there is no model fully suitable for all the traits, and the best model must be selected according to the accuracy and timeliness of prediction.

Key words: sika deer, growth-related traits, heritability, genome selection, accuracy