Table 1.
Performance comparison between the LoopIng, DisGro (DG) and LoopWeaver (LW) methods on the CASP10 dataset
| Loop length (# of cases) | LoopIng (a) (Å) |
DG (b) (Å) |
LW (c) (Å) |
Prediction ≤ 1 Å (d) (%) |
LoopIng < DG (e) (%) |
LoopIng < LW (f) (%) |
|||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Mean | SD | Mean | SD | Mean | SD | LoopIng | DG | LW | LoopIng < DG (I) | (DG – LoopIng) ≥ 1 Å (II) | LoopIng < LW (I) | (LW – LoopIng) ≥ 1 Å (II) | |
| 4 (79) | 0.74 | 0.63 | 1.43* | 0.68 | 0.93 | 0.64 | 70 | 27 | 47 | 75 | 38 | 52 | 16 |
| 5 (81) | 0.85 | 0.70 | 1.77* | 0.79 | 1.16* | 0.76 | 62 | 15 | 42 | 76 | 51 | 68 | 23 |
| 6(57) | 1.06 | 0.75 | 2.06* | 0.86 | 1.8* | 0.87 | 58 | 12 | 21 | 83 | 52 | 79 | 38 |
| 7(51) | 1.6 | 0.88 | 2.05* | 0.83 | 2.5* | 0.70 | 29 | 9 | 7 | 61 | 32 | 81 | 39 |
| 8(35) | 1.88 | 0.98 | 2.47* | 0.88 | 2.6* | 0.74 | 24 | 4 | 4 | 72 | 40 | 76 | 40 |
| 9(30) | 1.7 | 1.2 | 2.63* | 1.00 | 3.2* | 0.62 | 45 | 5 | 0 | 60 | 50 | 90 | 45 |
| 10(19) | 2.4 | 1.23 | 3.45* | 1.62 | 3.4* | 0.85 | 22 | 0 | 0 | 76 | 45 | 67 | 56 |
| 11(19) | 2.38 | 1.4 | 3.2 | 1.9 | 3.0 | 1.02 | 33 | 0 | 11 | 76 | 56 | 78 | 22 |
| 12(23) | 1.8 | 1.65 | 3.55* | 1.6 | 2.69 | 1.4 | 46 | 0 | 15 | 77 | 69 | 77 | 46 |
| 13(13) | 3.1 | 1.39 | 3.9 | 1.8 | 3.2 | 1.75 | 0 | 0 | 0 | 60 | 0 | 53 | 33 |
| Overall (407) | 1.29 | 1.07 | 2.09 | 1.44 | 1.98 | 1.71 | 51 | 14 | 25 | 73 | 44 | 71 | 31 |
Asterisks indicate a statistically significant difference (95% confidence level) with respect to the LoopIng method based on an unpaired t-test. (a, b, c) Mean RMSD and Standard Deviation for LoopIng, DG, and LW respectively. (d) Percentage of cases where LoopIng, DG and LW were able to give a prediction closer than 1 Å with respect to the native loop. (e, f) percentage of cases where LoopIng was more accurate (LoopIng < DG) and significantly better (ΔRMSD ≥ 1 Å) compared to DG and LW respectively.