V-Phaser 2: variant inference for viral populations

BMC Genomics

Table 3 Variant inference results of V-Phaser, QuRe and V-Phaser 2 on three datasets

Data		Method	TP	FP	FN	TN	Sensitivity	Specificity	Run time (min)	Memory (G)
		V-Phaser	110	116	4	694	96.49%	85.56%	34.3	12.51
	454	QuRe	59	19	55	791	51.75%	97.65%	6.5	7.50
		V-Phaser 2	105	27	9	783	92.11%	96.67%	0.9	0.04
8-mix	Illumina	V-Phaser	-	-	-	-	-	-	>600.0	>100.00
	MiSeq	QuRe	87	84	27	9,516	76.31%	99.13%	206.3	11.00
		V-Phaser 2	106	40	8	9,560	92.98%	99.58%	36.1	0.73
		V-Phaser	194	180	49	2,711	79.84%	93.84%	120.6	18.20
24-mix	454	QuRe	124	201	119	2,690	51.03%	93.05%	19.5	7.80
		V-Phaser 2	196	61	48	2,829	80.33%	97.89%	2.4	0.14

The results show that V-Phaser 2 substantially reduces the run-time and memory usage compared to V-Phaser and QuRe; V-Phaser 2 has comparable sensitivity with V-Phaser, where both are better than QuRe; V-Phaser 2 has comparable specificity compared to QuRe, where both are better than V-Phaser. ‘-’ indicates the corresponding value was not measured. We terminated V-Phaser after it uses over 100G memory on the Illumina MiSeq data. For each performance measure, the best value is bolded.

ISSN: 1471-2164