Toward an accurate prediction of inter-residue distances in proteins using 2D recursive neural networks

BMC Bioinformatics

Table 1 Reconstruction of C _α -traces from native and non-native maps

Maps	RMSD [Å]	GDT_TS	TM-score
Binary	4.38 (0.90, 14.98)	0.72 (0.42, 0.96)	0.77 (0.29, 0.97)
Binary ± 3Å	4.05 (1.50, 12.44)	0.64 (0.36, 0.82)	0.74 (0.42, 0.90)
Binary ± 6Å	4.26 (2.54, 9.78)	0.53 (0.32, 0.67)	0.64 (0.29, 0.78)
4-Class	1.04 (0.47, 6.90)	0.94 (0.73, 1.00)	0.95 (0.79, 0.98)
4-Class ± 3Å	1.41 (0.88, 6.80)	0.85 (0.67, 0.93)	0.90 (0.72, 0.96)
4-Class ± 6Å	2.25 (1.53, 4.08)	0.70 (0.56, 0.81)	0.81 (0.57, 0.88)
Distance	0.48 (0.22, 0.87)	0.99 (0.94, 1.00)	0.99 (0.94, 0.998)
Distance ± 3Å	0.96 (0.66, 1.46)	0.92 (0.85, 0.98)	0.94 (0.73, 0.99)
Distance ± 6Å	1.62 (1.03, 4.20)	0.81 (0.57, 0.88)	0.87 (0.48, 0.96)

The reconstruction of C_α-traces derived from binary contact maps, 4-class contact maps and distance maps. The native maps and the maps with a random error of 3Å and 6Å are used with the basic reconstruction protocol. Average RMSD [Å], GDT_TS [fraction] and TM-score, along with their range (min, max) are reported using the CASP7 targets.

ISSN: 1471-2105