Figure 7From: EnzML: multi-label prediction of enzyme classes using InterPro signaturesReannotation of proteomes. Left panel: reannotation of individual species proteomes. The classifier is trained on the SwissProt⋈KEGG dataset (minus the species to be predicted) and then used to predict each species proteome. The x axis (accuracy and recall) starts at 65%. Right panel: reannotation of the entire SwissProt⋈KEGG dataset starting from a single species proteome. The classifier is trained on a single proteome and then used to predict all the other species. The dashed line at 45% represents the baseline of subset accuracy than would be obtained if all proteins were simply classified as non-enzyme. There are no standard deviation bars since no randomisation is involved: each value represents one experiment (one species excluded or all other species excluded)Back to article page