Table 3.

Hydrophobic lengths calculated using HDGBv3 and HDGBvdW models.

PDB ID	Exp.	IMM1	GBIM	HDGBv3	HDGBvdW	OPM	PDBTM
1hxx	20.3±1	10	23	17	19	24.0 ± 0.8	19.0
1fep	22.6	15	30	23	23	24.3 ± 1.1	20.3
1nqe	20	12	26	20	21	23.4 ± 1.0	20.0
1qj8	28	18	27	22	24	23.6 ± 2.8	28.5
1qfg	23	18	30	21	22	24.7 ± 1.0	25.5
1qd6	15–20	17	29	21	22	23.9 ± 1.0	22.5
1qjp	23	9	28	22	23	25.4 ± 1.5	20.0
1m0l	32	25	29	31	30	31.8 ± 1.1	24.5
1gzm	30	28	35	30	30	32.2 ± 1.5	32.5
1rzh	28	25	29	28	28	31.8 ± 0.8	24.0
1v55	27±1	27	32	20	22	28.0 ± 0.6	36.0
1r3j	34±1	25	34	30	30	34.8 ± 1.2	34.5
2oar	24±1	27	17	20	22	36.1 ± 2.2	32.5
1yce	34	31	40	36	36	37.0 ± 0.5	35.5
2cfp	31	25	35	31	31	31.1 ± 1.4	29.5
avge	26.5	20.8	29.6	24.8	25.5	28.8	27.0
R2		0.61	0.43	0.78	0.82	0.58	0.51
slope		1.11	0.72	1.00	0.89	0.76	0.88
95% CIa		0.48	0.45	0.29	0.22	0.35	0.47
pb		5.4·10−4	8.2·10−3	1.3·10−5	3.2·10−6	9.7·10−4	2.7·10−3
MRc		−5.66	3.14	−1.66	−0.93	2.35	0.53
MSEd		50.32	27.67	9.27	5.12	16.52	17.68
1p4t	-	18	28	24	24	24.9 ± 2.4	32.5
2qom	-	13	35	16	24	25.1 ± 1.6	27.5
2x9k	-	20	28	22	25	24.7 ± 1.2	31.5
3aeh	-	9	30	17	22	25.2 ± 1.1	30.0
3dzm	-	23	35	29	32	28.5 ± 2.6	21.0
3syb	-	13	28	19	19	23.6 ± 0.9	21.5
4afk	-	15	26	22	22	24.8 ± 0.9	28.5
4d5b	-	14	28	24	23	23.5 ± 1.0	33.9
4fqe	-	9	23	12	18	22.2 ± 1.5	25.0
4rl8	-	16	28	23	24	23.4 ± 1.1	32.5
1okc	-	22	28	25	25	29.5 ± 1.7	33.0
3wbn	-	23	27	26	25	31.8 ± 1.4	30.5
3wfd	-	26	34	22	24	31.7 ± 1.6	24.5
3wxv	-	24	29	24	29	32.8 ± 1.2	30.5
4hyj	-	21	21	25	25	30.0 ± 1.6	36.0
4m48	-	23	28	24	20	30.8 ± 1.6	27.5
4n6h	-	29	15	37	36	34.0 ± 2.3	36.0
4uc2	-	23	29	28	29	30.4 ± 0.9	31.0
4x5n	-	25	25	28	28	36.8 ± 1.1	32.5
4yl3	-	11	38	30	29	29.8 ± 0.7	26.5
avgf		19.7	28.8	24.3	25.3	28.4	28.5

Comparison with the experimental values (exp) and calculated values by OPM (PPM2 version) and PDBTM servers (PDBTM version: 2017-06-16) were presented. The OPM and PDBTM values were taken from the reported data for each PDB ID at the respective websites. Experimental hydrophobic lengths were taken from different studies; for 1fep, it is derived from the depth analysis; for 1nqe and 2cfp, it is calculated as the distances of Cα atoms of the membrane spanning residues along the z-direction (the membrane spanning region of 1nqe and the first principle axis of 2cfp were aligned along the z-axis for this calculation); for 1hxx, 1v55, 1r3j and 2oar, matching bilayer widths were extracted from the study of Lewis and Engelman⁶⁹ along with the experimental errors. For 1nqe and 1gzm, 10 Å was subtracted from head to head distances to obtain hydrophobic length.⁵⁸ For 1qd6, a range was reported and 20 Å was used as the experimental value for the statistical analysis. Because our protocol is deterministic, no statistical errors can be determined if only a single input structure is considered. R² and the slope were calculated from linear regression with respect to available experimental values.

^aconfidence interval for slope;

^bp-value from the F-test for significance of a linear fit;

^cmean residuals;

^dmean squared error between predicted and experimental values;

^eaverage over first set for which experimental values are available;

^faverage over all proteins.