Table 1.

Performance comparison on simulated virus population:

	ViVan	LoFreq	VPhaser2	SAMTools pileup
True positives	231	219	196	244
False positives	4	1	2	1399
True negatives	31918	31921	31920	30523
False negatives	13	25	48	0
sensitivity	0.947	0.898	0.803	1.000
specificity	1.000	1.000	1.000	0.956
PPV	0.983	0.995	0.990	0.149
NPV	0.999	0.998	0.995	1.000
F1-Score	0.965	0.944	0.887	0.259

Simulated DENV population using six sequenced clinical samples sub-sampled at various rates was used to demonstrate ViVan’s performance in the context of real sequencing data

Because the coverage in this data set was low (100×) ViVan’s PPV was the lowest. However, ViVan demonstrated the highest sensitivity out of the three methods, identifying all but one of the variants found in the other. This dataset also highlights the need for sequencing-error aware methods in order to reduce false positive calls such as the ones produced by naïve pileup.