. 2012 Oct 3;34(1):57–65. doi: 10.1002/humu.22225

Table 2.

Performance of Computational Prediction Methods using the VariBench Benchmarking Dataset

	tp	fp	tn	fn	Accuracy^a	Precision^a	Specificity^a	Sensitivity^a	NVP^a	MCC^a
Theoretical/unweighted computational prediction methods
SIFT	10,464	4,856	12,188	7,433	0.65	0.64	0.62	0.68	0.66	0.30
PolyPhen 1^b	10,093	9,185	17,669	3,199	0.69	0.77	0.85	0.52	0.64	0.39
PolyPhen 1^c	14,285	4,993	13,671	7,197	0.70	0.68	0.66	0.74	0.72	0.40
PANTHER	9,689	2,859	8,676	2,797	0.76	0.76	0.76	0.77	0.77	0.53
FATHMM (unweighted)	11,561	4,839	16,257	7,707	0.69	0.72	0.77	0.60	0.66	0.38
Trained/weighted computational prediction methods
PolyPhen 2^b	13,807	5,102	13,863	6,010	0.71	0.71	0.70	0.73	0.72	0.43
PolyPhen 2^c	16,206	2,703	10,199	9,674	0.69	0.64	0.51	0.86	0.78	0.39
PhD-SNP	11,900	6,896	16,788	4,377	0.71	0.75	0.79	0.63	0.68	0.43
SNPs&GO	13,736	5,487	17,028	1,382	0.82	0.90	0.92	0.71	0.76	0.65
nsSNPAnalyzer	4,360	2,778	1,319	943	0.60	0.59	0.58	0.61	0.60	0.19
SNAP	16,000	2,146	8,190	6,387	0.72	0.67	0.56	0.88	0.83	0.47
MutPred	13,829	2,507	15,891	4,557	0.81	0.79	0.78	0.85	0.84	0.63
FATHMM (weighted)	14,231	1,633	10,146	2,336	0.86	0.86	0.86	0.86	0.86	0.72

tp, fp, tn, fn refer to the number of true positives, false positives, true negatives, and false negatives, respectively.

Accuracy, Precision, Specificity, Sensitivity, NVP, and MCC are calculated from normalized numbers.

“Probably Pathogenic” predictions classed as disease causing.

“Probably Pathogenic” predictions classed as functionally neutral.