Skip to main content
Fig. 2 | BMC Genomics

Fig. 2

From: Analysis of merged whole blood transcriptomic datasets to identify circulating molecular biomarkers of feed efficiency in growing pigs

Fig. 2

Regression analysis of the relationship between observed and predicted FCR. A predictive model to identify the most important annotated expressing probes able to predict feed-conversion-ratio (FCR) was built from the whole blood transcriptome merged from three independent experiments, and using a Gradient TreeNet Boosting (GTB) algorithm. Randomly selected bootstrap pig samples (n = 74) were used for learning, whereas the remaining samples (n = 74) were used for validation. Iterative steps led to retain a set of 50 very important variables. The graph was then computed between observed and predicted FCR values. Accuracy of the prediction was estimated by using R squared (R2) and root mean square error of prediction (RMSEP). Pigs considered in the study were from two divergent selection lines for residual feed intake (RFI), a measure of net feed efficiency. The red square represents pigs of the high RFI line, and the blue dot represents pigs of the low RFI group. No specific bias in prediction was observed due to RFI line

Back to article page