Fig. 1
From: EnsembleFam: towards more accurate protein family prediction in the twilight zone
Homology between training and test set of COG dataset. The bars indicate the fraction of test data having identity less than or equal to the indicated value on the x-axis. For each fold of the dataset, the homology is calculated for test sequence against the training sequences and the seed sequences used to build pHMM feature models. For each identity percentage, the three different bars indicate the average of 3-fold of the three different subsets of COG dataset