A machine learning approach for viral genome classification

BMC Bioinformatics

Table 2 Evaluation of HIV-1 classification with CASTOR

	Classification	# of classes	# of instances	[min - max] instances/class	TPR	FPR	F-measure	Classifier ID
Complete genomes	Groups (M, N, O and P)	4	76	[4 – 32]	1.000	0.000	1.000	PMVHIVGC01
	Pure subtypes	6	189	[30 – 36]	0.995	0.001	0.995	PMVHIVGC02
	CRFs	12	234	[10 – 30]	1.000	0.000	1.000	PMVHIVGC03
	Pure subtypes and CRFs	18	423	[10 – 36]	0.981	0.001	0.981	PMVHIVGC04
	Pure subtypes vs CRFs	2	200	[100 – 100]	0.795	0.205	0.795	PMVHIVGC05
pol fragments	Groups (M, N, O and P)	4	94	[4 – 45]	1.000	0.000	1.000	PMVHIVPL01
	Pure subtypes	6	1800	[300 – 300]	0.983	0.003	0.983	PMVHIVPL02
	CRFs	16	480	[30 – 30]	0.971	0.002	0.971	PMVHIVPL03
	CRFs	6	1200	[200 – 200]	0.993	0.001	0.993	PMVHIVPL04
	Pure subtypes and CRFs	23	690	[30 – 30]	0.920	0.004	0.919	PMVHIVPL05
	Pure subtypes and CRFs	12	2400	[200 – 200]	0.962	0.003	0.962	PMVHIVPL06
	Pure subtypes vs CRFs	2	200	[100 – 100]	0.885	0.115	0.885	PMVHIVPL07

This table contains the TPR, FPR and F-measure of 12 HIV-1 classifications obtained with 10-fold cross-validation analysis. For each classification, the number of corresponding classes and instances are given. The range [min-max] indicates the interval of instance frequencies per class used during the training of each model. The column Classifier ID contains the corresponding models available in CASTOR platform

ISSN: 1471-2105