EnsembleFam: towards more accurate protein family prediction in the twilight zone

BMC Bioinformatics

Table 5 Prediction accuracy comparison of different methods on the twilight zone proteins

Method	Sub-subfamily	Sub-family	Family
Identity: \(0 < x \le 30\)
pHMM	5.51	11.76	39.80
DeepFam	5.53	16.88	61.44
EnsembleFam	30.92	45.15	65.45
Identity: \(30 < x \le 40\)
pHMM	14.74	21.72	85.37
DeepFam	22.38	37.18	73.40
EnsembleFam	30.38	49.65	65.46

Best results are highlighted in bold font. For pHMM and EnsembleFam, we removed the predictions where the number of prediction is \(> 5\) and considered them as wrong prediction. For others, where the number of prediction is \(\le 5\) and the true label is included within the predicted one, we consider as correct. For DeepFam, as it only predicts one label, if the predicted label is the same as true label then we consider it as correct. EnsembleFam outperforms other two method in almost all cases

ISSN: 1471-2105