Using machine learning to improve the accuracy of genomic prediction of reproduction traits in pigs

• DOI: 10.1186/s40104-022-00708-0
• Publication Date: 2022-05-17T00:04:17Z
• Authors: Xue Wang; Shaolei Shi; Guijiang Wang; Wenxue Luo; Xia Wei; Ao Qiu; Fei Luo; Xiangdong Ding
• Keywords: AdaBoost

Recently, machine learning (ML) has become attractive in genomic prediction, but its superiority prediction over conventional (ss) GBLUP methods and the choice of optimal ML need to be investigated.In this study, 2566 Chinese Yorkshire pigs with reproduction trait records were genotyped GenoBaits Porcine SNP 50 K PorcineSNP50 panels. Four methods, including support vector regression (SVR), kernel ridge (KRR), random forest (RF) Adaboost.R2 implemented. Through 20 replicates fivefold cross-validation (CV) one for younger individuals, utility was explored. In CV, compared BLUP (GBLUP), single-step (ssGBLUP) Bayesian method BayesHE, significantly outperformed these methods. improved accuracy GBLUP, ssGBLUP, BayesHE by 19.3%, 15.0% 20.8%, respectively. addition, yielded smaller mean squared error (MSE) absolute (MAE) all scenarios. ssGBLUP an improvement 3.8% on average that close GBLUP. RF Adaboost.R2_KRR performed better than while comparably RF, slightly higher lower MSE total number piglets born, born alive, ssGBLUP. Among consistently well our study. Our findings also demonstrated hyperparameters are useful After tuning CV predicting outcomes 14.3% 21.8% those using default hyperparameters, respectively.Our had overall performance selection could new options prediction. is necessary The depend character traits, datasets etc.