. 2021 Jun 11;30(9):1833–1853. doi: 10.1002/pro.4136

TABLE 3.

Statistical analyses of various algorithms' abilities to discriminate Neutral (N), Rheostat (R), and Toggle (T) positions

Analysis	TC (N)	TC (R)	TC (T)^a	FC T\|N	FC R\|N	FC R\|T^b	FC N\|T	FC T\|R	FC N\|R	Thresholds:C1, C2
LacI^c
ConSurf (continuous)	0.8276	0.5566	0.6323	0.0299	0.1425	0.2794	0.0882	0.2082	0.2351	−1.008, 0.1661
ConSurf (discrete)^d	0.6964	0.5421	0.7985	0.0619	0.2417	0.1657	0.0357	0.2157	0.2421	4.6407, 7.8375
Combination	0.7981	0.5518	0.6351	0.1672	0.0348	0.3024	0.0625	0.2316	0.2166	1.1381, 2.1954
fuNTRp	0.6842	0.3307	0.7381	0.0789	0.2368	0.1667	0.0952	0.4016	0.2677	N/A
hLPYK
fuNTRp	0.667	0.375	0.400	0	0.3333	0.400	0.200	0.3125	0.3125	N/A

^{^a}

TC (i): Given that the truth was i, what was the probability that a position was correctly classified as i, with i corresponding to N, R, and T, as indicated. The threshold values used to carry out these calculations are in the last column.

^{^b}

FC (i|j): Given that the truth was j, what was the probability that a position was misclassified as i, with i and j corresponding to N, R, and T, as indicated. The threshold values used to carry out these calculations are in the last column.

^{^c}

Results for other bioinformatic analyses of LacI scores are in Tables S4 and S5.

^{^d}

The threshold values shown for discrete ConSurf scores were determined after applying kernel estimates. In practice, since the nature of these ConSurf scores is discrete, thresholds are rounded to 4.5 and 7.5.