Table 2.

Performance of the LoopIng method on the FREAD benchmark

Loop length	Original FREAD		LoopIng		LEAP
	Mean	SD	Mean	SD	Mean	SD
4	1.29*	1.14	0.61	0.55	0.39*	0.23
5	2.19*	2.02	0.68	0.52	0.40*	0.27
6	1.79*	1.37	1.01	0.63	0.49*	0.33
7	2.53*	2.34	1.26	0.9	0.69*	0.38
8	2.88*	2.37	1.47	1.07	0.68*	0.56
9	3.08*	2.60	1.71	1.23	0.93*	0.69
10	4.25*	3.58	1.90	1.34	1.44	0.84
11	4.55*	3.63	1.93	1.48	2.24	1.08
12	3.99*	3.88	2.20	1.70	3.14	2.52
13	5.54*	4.25	2.39	1.85	2.91	2.62
14	6.07*	4.36	2.53	3.03	4.44*	3.70
15	6.41*	5.05	3.05	3.00	4.58	4.16
16	7.50*	6.15	2.82	3.17	4.90*	4.43
17	7.84*	5.27	3.03	3.15	5.66*	5.50
18	5.48	5.64	3.86	3.47	6.53*	6.30
19	7.67*	5.27	3.89	3.51	5.87	4.64
20	7.64*	6.43	3.91	3.49	8.21*	7.82

For each length range the number of tested loops is 30. The columns report the average and standard deviation RMSD values measured between the model and native loop backbone conformations. Asterisks indicate statistically significant differences (95% confidence level) based on an unpaired t-test with respect to the LoopIng model. Underlined values represent the best results between LoopIng and LEAP. The values reported in the FREAD and LEAP columns are taken from Choi and Deane (2010) and Liang et al. (2014), respectively.