Performance of Web tools for predicting changes in protein stability caused by mutations

BMC Bioinformatics

Table 5 Consensus of the predictors

	Monomeric proteins		Multimeric proteins
	3/5 methods	2/3 methods^a	3/5 methods	2/3 methods^{b, c}
Values calculated for the full dataset
True negative	426	407	134	133
False positive	17	36	8	9
True positive	63	73	22	24
False negative	39	29	23	21
Accuracy	0.90	0.88	0.83	0.84
True negative rate (specificity)	0.96	0.92	0.94	0.94
True positive rate (sensitivity)	0.62	0.72	0.49	0.53
Positive predictive value (precision)	0.79	0.67	0.73	0.72
Negative predictive value	0.92	0.93	0.85	0.86
MCC	0.64	0.62	0.50	0.53
Values calculated for the balanced dataset
True negative	92	87	36	36
False positive	10	15	2	2
True positive	63	73	21	21
False negative	39	29	24	24
Accuracy	0.76	0.78	0.69	0.69
True negative rate (specificity)	0.90	0.85	0.95	0.95
True positive rate (sensitivity)	0.62	0.72	0.47	0.47
Positive predictive value (precision)	0.86	0.83	0.91	0.91
Negative predictive value	0.70	0.75	0.60	0.60
MCC	0.54	0.57	0.46	0.46

True negative and true positive values have been considered as those predictions that correctly found a negative and a positive sign for destabilizing and stabilizing mutations, respectively. Data have been reported for mutations causing a ΔΔG energy variation outside the range of the experimental error, in the full and the balanced datasets of both monomeric and multimeric proteins
^aResults obtained using DynaMut, DUET and INPS-MD
^b For the full dataset, results obtained using PoPMuSiC, DUET and MAESTROweb
^c For the balanced dataset, results obtained using PoPMuSiC, DUET and INPS-MD

ISSN: 1471-2105