. 2022 Nov 14;14(1):2143006. doi: 10.1080/19420862.2022.2143006

Table 1.

Performance analysis for the physics-based classifier model using different combinations of the descriptors: (D1) backbone dihedral conformation of the n + 1 residue, (D2) side-chain dihedral conformation of Asn/Asp residue, (D3) fraction of time the Asn/Asp residue remains solvent accessible. Combinations are represented by DXX where the X represents the descriptors.

Type	TP*	FN*	FP*	TN*	Accuracy	Accuracy^†	Precision	Recall	F₁ Score	F_0.5 Score	MCC^‡	CKS^§
ASN-D1	20	3	173	345	0.67	0.77	0.95	0.67	0.77	0.87	0.23	0.12
ASN-D2	21	2	301	217	0.44	0.67	0.95	0.44	0.57	0.75	0.14	0.05
ASN-D3	23	0	316	202	0.42	0.69	0.96	0.42	0.54	0.73	0.16	0.05
ASN-D12	19	4	130	388	0.75	0.79	0.95	0.75	0.83	0.90	0.26	0.16
ASN-D13	20	3	92	426	0.82	0.85	0.96	0.82	0.87	0.92	0.34	0.24
ASN-D23	21	2	183	335	0.66	0.78	0.96	0.66	0.76	0.86	0.23	0.12
ASN-naïve	0	23	0	518	0.96	0.50	0.92	0.96	0.94	0.92	0.00	0.00
ASN-D123	19	4	62	456	0.88	0.85	0.96	0.88	0.91	0.94	0.40	0.32
ASP-D1	22	2	186	288	0.62	0.76	0.95	0.62	0.73	0.84	0.23	0.11
ASP-D2	22	2	318	156	0.36	0.62	0.94	0.36	0.48	0.68	0.11	0.03
ASP-D3	21	3	312	162	0.37	0.61	0.94	0.37	0.49	0.68	0.10	0.03
ASP-D12	20	4	118	356	0.76	0.79	0.95	0.76	0.82	0.89	0.28	0.18
ASP-D13	20	4	133	341	0.72	0.78	0.95	0.72	0.80	0.88	0.26	0.16
ASP-D23	20	4	205	269	0.58	0.70	0.94	0.58	0.69	0.82	0.17	0.08
ASP-naïve	0	24	0	474	0.95	0.50	0.91	0.95	0.93	0.91	0.00	0.00
ASP-D123	19	5	77	397	0.84	0.81	0.95	0.84	0.88	0.92	0.34	0.26
ALL-D1	42	5	359	633	0.65	0.77	0.95	0.65	0.75	0.86	0.23	0.12
ALL-D2	43	4	619	373	0.40	0.65	0.95	0.40	0.53	0.72	0.13	0.04
ALL-D3	44	3	628	364	0.39	0.65	0.95	0.39	0.52	0.71	0.13	0.04
ALL-D12	39	8	248	744	0.75	0.79	0.95	0.75	0.83	0.90	0.27	0.17
ALL-D13	40	7	225	767	0.78	0.81	0.95	0.78	0.84	0.90	0.30	0.19
ALL-D23	41	6	388	604	0.62	0.74	0.95	0.62	0.73	0.85	0.20	0.10
ALL-naïve	0	47	0	992	0.95	0.50	0.91	0.95	0.93	0.92	0.00	0.00
ALL-D123	38	9	139	853	0.86	0.83	0.95	0.86	0.89	0.93	0.37	0.29

*TP – True positive, FN – False Negative, FP – False positive and TN – True Negative. † Balanced accuracy calculated with weighted distribution of positives and negatives. ‡ Matthew’s Correlation Coefficient ranges from −1.0 to +1.0, where the negative values indicate a bad predictor. § Cohen Kappa Score ranges from −1.0 to +1.0, similar to the MCC statistic but a better indicator to compare an imbalanced dataset.