Assessment of bagging GBLUP for whole‐genome prediction of broiler chicken traits

摘要: Summary Bootstrap aggregation (bagging) is a resampling method known to produce more accurate predictions when predictors are unstable or the number of markers much larger than sample size, because variance reduction capabilities. The purpose this study was compare genomic best linear unbiased prediction (GBLUP) with bootstrap aggregated sampling GBLUP (Bagged GBLUP, BGBLUP) in terms accuracy. We used 600 K Affymetrix platform 1351 birds genotyped and phenotyped for three traits broiler chickens; body weight, ultrasound measurement breast muscle hen house egg production. predictive performance versus BGBLUP evaluated different scenarios consisting including excluding TOP 20 from standard genome-wide association (GWAS) as fixed effects model, varying training sizes allelic frequency bins. Predictive assessed via five replications threefold cross-validation using correlation between observed predicted values, mean-squared error. overfitted set data, delivered better ability testing sets. Treating GWAS into model improved accuracy added advantages over GBLUP. at allele bins similar. In general, results confirm that can be valuable enhancing genome-enabled complex traits.