Frequently asked questions
1. My LocScale map still looks noisy. What can I do to improve it?
(a) Make sure atomic B-factors are properly refined. The distribution of atomic B-factors should roughly follow the local resolution estimate for the map.
(b) By default LocScale generates the reference map up to Nyquist frequency. In case the average resolution is significantly lower, this may lead to overemphasizing noise terms in the scaling procedure. You can limit the resolution of the reference map using the -dmin
keyword in the prepare_locscale_input.py
script, or low-pass filter your LocScale volume at a frequency slightly lower than Nyquist.
2. LocScale does not seem to have any effect on my map. What could be the reason?
(a) Please make sure your atomic B-factors have been refined prior to scaling. LocScale uses the locally estimated amplitude falloff of the reference model to sharpen the experimental map.
(b) Make sure your atomic model (and reference map) overlap with mask and experimental map. Sometimes models are shifted by refinement programs; be sure to revert those shifts prior to generate reference maps for scaling.
(c) Make sure your original map is provided unfiltered. Frequencies filtered prior to scaling cannot be recovered during the scaling procedure.
3. It seems my map is inverted after running LocScale
(a) Currently, LocScale uses cctbx libraries to generate the reference map and the axis convention differs from that used in SPARX. Please be sure to permute the axis order prior to scaling. See here for instructions.
4. Model map generation stops with error "Gaussian not defined for scattering_type "O1-/N1+"
cctbx does not currently support electron form factors for charges atoms. Please see here if you want to proceed with charges, or use prepare_locscale_input.py
with the -s/--strip_charge
flag to force neutral atoms for model map generation.