Table 1.

Prediction results over the top 10 models.

Subset	Method	CASP11	CASP12	HOMSTRAD	<Å (%)	<2Å (%)
Commonai	best	2.18	2.31	1.65	24	66
	DaReUS-Loop	2.58	2.87	2.38	19	47
	Rosetta NGK	2.96	3.34	2.71	12	35
	GalaxyLoop-PS2	2.83	3.23	2.96	13	36
Commondb	DaReUS-Loop	2.79	3.21	2.53	14	44
	LoopIng	4.35	4.20	4.50	8	16
	Sphinx	3.71	3.94	2.63	12	37
CommonHCai	best	1.43	1.63	1.65	28	76
	DaReUS-Loop	1.91	2.30	2.38	22	54
	Rosetta NGK	2.59	2.99	2.71	14	38
	GalaxyLoop-PS2	2.34	2.88	2.96	15	41
CommonHCdb	DaReUS-Loop	2.05	2.25	2.53	17	53
	LoopIng	3.66	3.53	4.50	10	20
	Sphinx	2.90	3.19	2.63	14	42

Average flanked RMSD (Å) are reported for the CASP11, CASP12 and HOMSTRAD test sets, over the Common and CommonHC subsets. Comparison is between DaReUS-Loop, ab initio (_ai) methods (Rosetta NGK and GalaxyLoop-PS2) and data-based (_db) methods (LoopIng and Sphinx). All the RMSD values reported in this table correspond to the best flanked RMSD (Å) over 10 models. The “best” row shows the best candidate loop identified by DaReUS-Loop, before applying the filters or the top 10 selection. For Common_ai and CommonHC_ai (resp. Common_db and CommonHC_db), the flanked RMSDs are calculated using flanks of 4 (reps. 2) amino acids. The percentage of highly accurate predictions (<1Å and <2Å) is also reported. Bold values correspond to the best values among all the methods.