Table 3. Performance of the random forests models on validation sets V^60P+45N and V^60P+60N*.

Data	Model	Sensi tivity	Speci ficity	Accu racy	MCC
V60P+45N	RFcompo	86.7	80.0	83.8	0.67
	RFcompo+agg	88.3	82.2	85.7	0.71
	RFcompo+structure	90.0	86.7	88.6	0.77
	RFcompo+structure +agg	91.7	86.7	89.5	0.79
	RFcompo#	93.3	48.9	74.3	0.48
	RFcompo+agg#	91.7	51.1	74.3	0.48
	RFcompo+structure#	93.3	42.2	71.4	0.43
	RFcompo+structure +agg#	93.3	48.9	74.3	0.48
	RFphysico	93.3	77.8	86.7	0.73
	RFphysico+agg	90.0	82.2	86.7	0.73
	RFphysico+structure	91.7	82.2	87.6	0.75
	RFphysico+structure +agg	91.7	82.2	87.6	0.75
	RFphysico#	90.0	53.3	74.3	0.48
	RFphysico+agg#	88.3	53.3	73.3	0.45
	RFphysico+structure#	90.0	40.0	68.6	0.35
	RFphysico+structure +agg#	95.0	48.9	75.2	0.51
V60P+60N*	RFcompo	86.7	56.7	71.7	0.45
	RFcompo+agg	88.3	56.7	72.5	0.47
	RFcompo+structure	90.0	60.0	75.0	0.52
	RFcompo+structure +agg	91.7	56.7	74.2	0.52
	RFcompo#	93.3	93.3	93.3	0.87
	RFcompo+agg#	91.7	95.0	93.3	0.87
	RFcompo+structure#	93.3	88.3	90.8	0.82
	RFcompo+structure +agg#	93.3	86.7	90.0	0.80
	RFphysico	91.7	41.7	66.7	0.39
	RFphysico+agg	90.0	45.0	67.5	0.39
	RFphysico+structure	91.7	48.3	70.0	0.44
	RFphysico+structure +agg	91.7	48.3	70.0	0.44
	RFphysico#	90.0	95.0	92.5	0.85
	RFphysico+agg#	88.3	91.7	90.0	0.80
	RFphysico+structure#	90.0	88.3	89.2	0.78
	RFphysico+structure +agg#	95.0	86.7	90.8	0.82

V^60P+60N^* contained non-experimental peptides. The models trained by T^544P+544N^* were marked by the number sign #; the models trained by T^544P+407N had no marks.