Figure 1
From: PANDA: Protein function prediction using domain architecture and affinity propagation
The precision-recall plot of PANDA and other predictors. The predicted and real GO terms are not propagated to the root in the GO directed acyclic graph; and two GO terms are considered a match only if they are the same. The “Final-pool” is based on predictions pooled from all predictors before filtered by Z-score. The “Final-pool-filtered” is before applying GO term clustering using affinity propagation but after filtered by GO term Z-score. The “Final-pool-filtered-clustered” is after the clustering and second round of filtering that is based on clustering results. “Domain-architecture” is a red dot in the plot because we assign a constant confidence score to all predictions from it (see Methods for details).
