journal menu
![[Figure 3]](ba5259fig3mag.jpg)
|
|
Figure 3
Ensemble models consistently have less residual difference density than ligand-only models. All images show 2mFo − DFc maps contoured at 1.5σ (blue) and mFo − DFc maps contoured at ±3σ (green/red). First column, refinement with the ligand model only. Centre column, refinement of the crystal as an ensemble of states. Last column, refinement of the crystal as an ensemble of states with a degraded protein model (phase difference as indicated relative to the ensemble model). (a, d, g) Modelling the ligand but removing the ground state leads to difference density for the absent state, and in (d) the ligand moves into density for the ground state (r.m.s.d. of 0.20 in Fig. 4 ![[link]](../../../../../../logos/arrows/d_arr.gif) b). Arrows in (g) indicate the potential remodelling of the ligand by an overzealous modeller, as discussed in §3.3. (j) Removing the ground state for a high-occupancy ligand (refined value of 0.89) does not lead to discernible difference density. (b, e, h, k) Refinement of ensemble models explains all of the observed density, and ligands do not move from the fitted pose (confirmed by the validation plots in Fig. 4). (c, f, i, l) Refining with degraded phases leads to only minor visual differences in the model, except in (f) where the ligand moves significantly relative to the fitted pose. However, more striking is the disappearance of difference density, as expected, owing to the increase in the noise level of the map. |
![[Figure 3]](ba5259fig3.jpg)
 | STRUCTURAL BIOLOGY |
ISSN: 2059-7983
access
