Table 2. Comparative statistics for refinement of IAS and multipolar models.
Mt and Mr represent multipolar models with transferred and refined parameters (refinements ‘3’ and ‘5’ in Volkov et al., 2007 ▶). 〈B
nonH〉 is the mean value of the equivalent isotropic ADP calculated for non-H atoms. RBT is the rigid-bond-test value (the same as DMSDA, differences in mean-squared displacement amplitudes, in Volkov et al., 2007 ▶). R
work and R
free are the standard crystallographic R and R
free factors between experimental F
obs and model-based calculated structure-factor magnitudes F
model (Afonine et al., 2005 ▶) calculated as 
.
| Data set | Model | Ndata/Npar† | Rwork | Rfree | 〈BnonH〉 (Å2) | RBT (104 Å2) |
|---|---|---|---|---|---|---|
| YGG, low resolution | IAM‡ | 4.9 | 2.16 | — | — | 17.76 |
| Mt‡ | 6.2 | 1.22 | — | — | 12.85 | |
| IAM | 6.2 | 2.35 | 2.62 | 1.23 | 18.99 | |
| IAS | 4.0 | 1.57 | 2.00 | 1.05 | 12.23 | |
| YGG, high resolution | IAM‡ | 17.3 | 4.51 | — | — | 8.77 |
| Mt‡ | 21.9 | 3.66 | — | — | 7.38 | |
| Mr‡ | 10.6§ | 3.42 | — | — | 6.38 | |
| IAM | 21.9 | 4.57 | 4.72 | 1.04 | 8.62 | |
| IAS | 14.2 | 3.75 | 4.06 | 1.07 | 7.68 | |
| P2A4, low resolution | IAM‡ | 5.5 | 2.98 | — | — | 15.64 |
| Mt‡ | 7.1 | 1.84 | — | — | 7.09 | |
| IAM | 7.1 | 3.51 | 3.79 | 1.24 | 20.77 | |
| IAS | 4.5 | 2.45 | 3.27 | 1.07 | 16.77 | |
| P2A4, high resolution | IAM‡ | 46.7 | 3.44 | — | — | 3.67 |
| Mt‡ | 61.0 | 2.67 | — | — | 2.65 | |
| Mr‡ | 43.6§ | 2.53 | — | — | 3.09 | |
| IAM | 61.1 | 3.72 | 3.63 | 1.14 | 3.66 | |
| IAS | 38.1 | 3.06 | 3.23 | 1.14 | 4.79 | |
| Antifreeze protein | IAM | 18.6 | 12.77 | 15.37 | 7.84 | 208.4 |
| IAS | 14.3 | 11.76 | 14.44 | 7.40 | 195.7 | |
| Trypsin | IAM | 7.6 | 10.30 | 13.79 | 5.79 | 149.3 |
| IAS | 5.8 | 9.19 | 13.35 | 5.52 | 126.0 | |
| Phospholipase | IAM | 6.0 | 8.99 | 12.80 | 9.88 | 250.6 |
| IAS | 4.7 | 8.31 | 12.64 | 9.11 | 213.5 | |
| Scorpion toxin | IAM | 4.9 | 9.40 | 15.47 | 10.30 | 365.8 |
| IAS | 3.9 | 8.78 | 15.23 | 10.42 | 363.1 |
For multipolar refinement a number of parameters were fixed or linked by constraints. N par is the number of parameters at each step and does not include the number of parameters refined previously. In contrast to Volkov et al. (2007 ▶), in the current project the ratio N data/N par was calculated for the total number of refined parameters even when at each particular moment only a subset of them were refined; a direct comparison of this information with that reported in Volkov et al. (2007 ▶) is not straightforward.
Refined by Volkov et al. (2007 ▶); corresponding numbers are cited from there.
An estimate obtained if the same set of parameters were used for refinement at ‘high’ resolution.