Prediction accuracy of individual feed conversion ratio using group records in growing Duroc pigs
The objective was to evaluate the prediction accuracy of individual feed conversion ratio (iFCR) in six scenarios defined by the distribution of individual versus group feed conversion ratio (gFCR): 10%-50% and 90% of the population have iFCR instead of gFCR. Individual daily gain (ADG), individual backfat thickness (BF) and iFCR were available for 1384 Duroc pigs. To calculate iFCR, individual feed intake records were collected using electronic feeders, then gFCR was defined as the within pen iFCR average. The dataset was partitioned into seven subsets leaving out one different animal per pen every fold. Each subset was treated in each fold as the validation set (118), and a four-trait animal model (ADG, BF, iFCR and gFCR) was trained with the reaming records using EM-REML. Correlations between predicted and observed iFCR in the validation sets were averaged across replicates (sd). The prediction accuracies were similar when 90% or 50% iFCR were considered: 0.76(0.05) and 0.77(0.06) respectively. The prediction accuracy reduced to 0.33(0.11) when only 10% iFCR were available. When all records were treated as iFCR the accuracy of the prediction was 0.78(0.04). These results indicate that iFCR can be satisfactorily predicted including at least 50% individual records, having the rest of animals just gFRC. This is a clear indication of the value that easy-to-record feed intake traits could have within a breeding program for feed efficiency. Nonetheless, the consequence that these accuracies on phenotypes prediction might have on the prediction of breeding values has to be further explored.