Figure 5. Weak ligand identification remains straightforward when phases are degraded.

BAZ2B data sets were re-analysed using a deliberately sabotaged reference model, introducing a ∼30° phase error and increasing R_work and R_free by ∼12% for all dimple-refined data sets. Shown here is the weak hit (refined occupancy: 0.64) in data set x492, contoured for different maps as labelled: (a,b) 1.78 Å 2mF_o–DF_c (blue, 1σ) and mF_o–DF_c (green/red, ±3σ). (c,d) 1.79 Å event (blue, 2σ) and Z-maps (green/red, ±3). R_work/R_free are 0.18/0.21 and 0.30/0.32 for best and degraded phases, respectively. BDCs for best and degraded phases are 0.77 and 0.73, respectively. Whereas with standard maps, degraded phases remove all evidence of an unmodelled change, in PanDDA maps, ligand identification is no more difficult, even if the quality of the density is predictably reduced.