Table 3.

Comparison results of predictors for bitter/non-bitter and bitter/sweet prediction.

Classification Type	Model	Algorithm	Input	TN	FP	FN	TP	TPR	TNR	PRE	ACC	AP	F1	MCC	AUC
Bitter/Non-bitter	BoCNN	CNN	Molecular Figure	132	31	33	82	0.71	0.81	0.73	0.77	0.64	0.72	0.52	0.76
	BoMLP	MLP	Descriptors	119	44	33	82	0.71	0.73	0.65	0.72	0.58	0.68	0.44	0.72
	BoMLP	MLP	RDKFP	141	22	16	99	0.86	0.86	0.82	0.86	0.76	0.84	0.72	0.86
	BitterPredict	Adapt Boost	Descriptors	133	30	32	83	0.72	0.82	0.73	0.78	0.65	0.73	0.54	0.77
	VisualTaste	Random Forest	Descriptors	95	68	20	95	0.83	0.58	0.58	0.68	0.55	0.68	0.41	0.7
	BitterX	SVM	Descriptors	38	125	18	97	0.84	0.23	0.44	0.49	0.43	0.58	0.09	0.54
	HGNN	GNN	Atom and Bond Feature	121	42	12	103	0.9	0.74	0.71	0.81	0.68	0.79	0.63	0.82
	WLN	GNN	Atom and Bond Feature	140	23	14	101	0.88	0.86	0.81	0.87	0.77	0.84	0.73	0.87
	AttentiveFP	GNN	Atom and Bond Feature	150	13	24	91	0.79	0.92	0.88	0.87	0.78	0.83	0.72	0.86
	GraphSAGE	GNN	Atom Feature	146	17	18	97	0.84	0.9	0.85	0.87	0.78	0.85	0.74	0.87
Bitter/Sweet	BoCNN	CNN	Molecular Figure	114	25	29	86	0.75	0.82	0.78	0.79	0.69	0.76	0.57	0.78
	BoMLP	MLP	Descriptors	108	31	37	78	0.68	0.78	0.72	0.73	0.63	0.7	0.46	0.73
	BoMLP	MLP	RDKFP	126	13	23	92	0.8	0.91	0.88	0.86	0.79	0.84	0.71	0.85
	HGNN	GNN	Atom and Bond Feature	122	17	18	97	0.84	0.88	0.85	0.86	0.79	0.85	0.72	0.86
	WLN	GNN	Atom and Bond Feature	122	17	21	94	0.82	0.88	0.85	0.85	0.77	0.83	0.7	0.85
	AttentiveFP	GNN	Atom and Bond Feature	130	9	22	93	0.81	0.94	0.91	0.88	0.82	0.86	0.76	0.87
	GraphSAGE	GNN	Atom Feature	127	12	22	93	0.81	0.91	0.89	0.87	0.8	0.85	0.73	0.86