Table 3. Prediction performance comparison of Crysalis and other existing methods.

Experimental step	Method	AUC	MCC	ACC(%)	SPEC(%)	SENS(%)	PRE(%)
CLF	PredPPCrys I	0.711	0.296	65.33	63.58	66.50	73.16
	PredPPCrys II	0.725	0.322	66.54	65.56	67.20	74.44
	Crysalis I	0.731	0.332	66.98	66.60	67.22	75.56
	Crysalis II	0.759	0.365	68.34	69.95	68.34	76.85
MF	PPCPred	0.683	0.334	68.06	67.99	68.22	47.20
	PredPPCrys I	0.772	0.380	69.93	68.21	72.88	49.95
	PredPPCrys II	0.793	0.416	71.95	71.36	73.30	52.70
	Crysalis I	0.759	0.377	70.23	69.93	70.99	49.25
	Crysalis II	0.793	0.427	73.08	73.58	73.09	54.15
PF	PPCPred	0.612	0.183	58.83	62.23	57.08	74.57
	PredPPCrys I	0.800	0.460	74.83	70.52	77.02	83.77
	PredPPCrys II	0.872	0.579	79.73	81.43	78.86	89.31
	Crysalis I	0.796	0.436	73.87	67.80	73.87	82.47
	Crysalis II	0.801	0.411	71.22	72.67	70.48	83.48
CF	PPCPred	0.432	−0.014	55.23	32.21	61.24	75.53
	PredPPCrys I	0.712	0.280	67.05	67.65	66.91	89.42
	PredPPCrys II	0.735	0.175	69.47	68.89	69.50	97.80
	Crysalis I	0.739	0.281	65.50	70.59	64.23	89.80
	Crysalis II	0.752	0.337	62.57	85.29	56.93	93.97
CRYs	ParCrys	0.611	0.132	59.66	60.56	55.91	25.40
	OBScore	0.638	0.184	59.28	57.78	65.49	27.14
	CRYSTAP2	0.599	0.123	51.64	48.10	67.78	22.28
	XtalPred	—	0.224	65.04	65.61	62.51	29.31
	SVMCRYs	—	0.142	55.11	52.78	65.70	23.39
	PPCPred	0.704	0.254	63.63	62.09	70.67	29.03
	XtalPred-RF	—	0.205	60.94	59.67	66.41	27.56
	PredPPCrys I	0.770	0.326	69.65	69.30	71.13	35.23
	PredPPCrys II	0.838	0.428	76.04	76.21	75.30	42.64
	Crysalis I	0.788	0.339	71.00	70.89	71.41	35.50
	Crysalis II	0.838	0.435	76.27	76.28	76.20	42.84

Performance was evaluated using independent test datasets. Note that most methods (ParCrys, OBScore, CRYSTAP2, XtalPred, and SVMCRYs) only provide one-class prediction (CRYs) and PPCPred includes four-class (MF, PF, CF, and CRYs) predictors. Thus, we only compared the performance of these tools for valid classes. In the case of PredPPCrys, we compared its performance with Crysalis for all five classes.