LocScale performs local amplitude scaling based on a atomic reference structure.

LocScale uses rotationally averaged reference amplitudes from a refined atomic model to locally scale (sharpen) amplitudes of the Fourier transform of a cryo-EM potential map, keeping the phases constant.

For sharpening LocScale requires the unfiltered, unsharpened EM reconstruction (or half maps) along with a reference map computed from the atomic model. Optionally, a mask may be applied to speed up the computation.

Please follow these instructions to generate your model-based reference map with B-factor weighted structure factors using electron form factors.

Requires sparx , EMAN2 , (mpi4py)

*** Computes contrast-enhanced cryo-EM potential maps by local amplitude scaling ***

usage: locscale_mpi.py [-h] -em EM_MAP -mm MODEL_MAP -p APIX [-ma MASK]
                       [-w WINDOW_SIZE] -o OUTFILE [-mpi]

optional arguments:
  -h, --help            show this help message and exit
  -em EM_MAP, --em_map EM_MAP
                        Input filename EM map
  -mm MODEL_MAP, --model_map MODEL_MAP
                        Input filename PDB map
  -p APIX, --apix APIX  pixel size in Angstrom
  -ma MASK, --mask MASK
                        Input filename mask
  -w WINDOW_SIZE, --window_size WINDOW_SIZE
                        window size in pixel
  -o OUTFILE, --outfile OUTFILE
                        Output filename
  -mpi, --mpi           MPI version call by: "mpirun -np 4 python locscale.py
                        -em emmap.mrc -mm modmap.mrc -ma mask.mrc -p 1.0 -w 10
                        -mpi -o scaled.mrc"

Invoking locscale_mpi.py: The script needs to be invoked within the EMAN2/Sparx Python framework, i.e. if EMAN2 Python is the default Python framework:

python locscale_mpi.py [arguments]

Window size: We find that a window size of approximately (7 * average_map_resolution_in_Å)/pixel_size typically works well.

Computing time: We strongly recommend to run the parallelized version of LocScale using MPI. Computing LocScale maps on a a single node can be (very) slow. In comparison, computation of a LocScale map for EMD-5778 takes about 4 min on 50 CPUs using MPI.