Semi-supervised learning for somatic variant calling and peptide identification in personalized cancer immunotherapy

Table 1 Variant calling performance on sequencing datasets P1-P4 generated for four ovarian cancer patients

DataSet	Caller	# calls	TP	FP	TN	FN	TPR	PPV	F1
P1	SNVQ	515	48	114	0	0	100	29.63	45.71
P: 65	Strelka	435	48	34	80	0	100	58.54	73.85
U: 29758	2CP	65	41	8	106	7	85.42	83.67	84.54
Reseq: 162	PLATO	587	42	6	104	10	80.77	87.5	84
P2	SNVQ	597	147	65	3	2	98.66	69.34	81.44
P: 187	Strelka	619	149	59	9	0	100	71.63	83.47
U: 31210	2CP	187	133	39	29	16	89.26	77.33	82.87
Reseq: 217	PLATO	449	144	43	25	5	96.64	77.01	85.71
P3	SNVQ	629	61	13	8	1	98.39	82.43	89.71
P: 76	Strelka	306	62	11	10	0	100	84.93	91.85
U: 30289	2CP	76	57	1	20	5	91.94	98.28	95
Reseq: 83	PLATO	429	62	3	18	0	100	95.38	97.64
P4	SNVQ	482	48	23	87	2	96	67.61	79.34
P: 67	Strelka	380	50	94	16	0	100	34.72	51.55
U: 30176	2CP	67	45	2	108	5	90	95.74	92.78
Reseq: 160	PLATO	490	48	7	103	2	96	87.27	91.43

ISSN: 1471-2105