Table 5. Performances of unsupervised disulfide connectivity predictors when added to supervised Machine Learning methods.

Method	Number of Bonds
	2		3		4		5		Average
	Rb	Qp	Rb	Qp	Rb	Qp	Rb	Qp	Qp
Random	33	33	20	7	14	1	11	0.1	15
SVR	75	75	60	48	57	44	46	19	54
SVR+MIp+ICOV	76	76	63	55	68	51	59	32	59
skSVR	79	79	67	60	60	41	55	28	60
skSVR+Sephiroth	86	86	71	64	67	50	66	46	68
skSVR+PhyloCys	82	82	68	59	69	59	68	51	67
skSVR+Sephiroth+PhyloCys	87	87	69	61	74	62	70	49	69

Performances of different supervised Machine Learning-based methods on the PDBCYS dataset. SVR and SVR+MIp+ICOV scores are reported from [9]; Sephiroth performances have been obtained with 3 iter E-value = 10⁻² HHblits MSAs and are reported from [16], PhyloCys scores are obtained with 3 iter E-value = 10⁻⁵.