This is an old revision of the document!


Sharpen3D

Map sharpening allows to correct for the contrast loss at high resolution, resulting in better interpretable maps.

Use this module to sharpen and subsequently low-pass filter a 3d map. Sharpening helps in interpreting the map: dependent on resolution of the map domains, secondary structure elements or side-chains will be more clearly defined. Sharpening can be either performed using a standard curve from SAXS data (“Do not use custom experimental data”; generally more conservative) or using a reference curve from another 3d volume (“Use custom experimental data”), e.g. a theoretical density computed from an atomic model. In the latter case, the pixel size of the reference must be provided. Sharpening also increases the high-resolution noise, which may impede reliable interpretation, in particular in less-well resolved regions of the map. Therefore, for homogeneously well-resolved maps subsequent filtering can performed in global mode, i.e. every part of the map is low-pass filtered to the same resolution. Heterogeneously resolved maps should be filtered in local mode, i.e. individual regions are filtered according to the respective local resolution.

Parameters Description
Use custom experimental data - Do not use Use standard curve for sharpening
Use custom experimental data - Use Use reference curve from custom 3d volume for sharpening
→ Experimental sampling Pixel size of reference 3d volume in Å
Filtering mode - global Low-pass filter sharpened 3d volume everywhere to the same resolution level
→ Resolution level Value for global low-pass filtering in Å
Filtering mode - local Low-pass filter sub-regions of the 3d volume map according to local resolution
→ Kernel radius Edge-length of cubic sub-regions in pixels
→ Resolution threshold Lowest resolution to which sub-regions are low-pass filtered
Filtering mode - none Omit low-pass filtering of resulting sharpened map
Normalize Check this box to normalize the sharpened 3d volume to mean 0 and sigma 10.
Pixel size Pixel size of the 3d volume to be sharpened.
Input Description
3d volume 3d volume to be sharpened
Optional experimental data Custom 3d volume to be used as reference for sharpening
Resolution levels Local resolution values (“Resolution levels” output) from FourierShellCorrelation logic
ThirdInput Local resolution map (“Fourier shell correlation” output) from FourierShellCorrelation logic
Output Description
1d power spec of input 1d curve showing the 1d averaged power spectrum of the input 3d volume
1d power spec of output 1d curve showing the 1d averaged power spectrum of the sharpened 3d volume
Sharpened 3D Sharpened and possibly filtered 3d volume
New/Changed Header Values Description
pixelSize Pixel size in Å

Experimental Sampling [Å]

The pixel size of the input “Optional experimental data” [in Angstrom]. If no experimental data input is given, this value is not used.

Fill filter with zeros

Check this checkbox to fill the filter image with zeros at indexes that are outside the important diameter. Otherwise the filter image will be filled using constant extrapolation.

Normalize

Check this checkbox to normalize the corrected 3d volume to mean 0 and sigma 10.

Pixel size [Å]

The pixel size of the input “3d volume” [in Angstrom].

Resolution Cutoff [px]

Cut off resolution in pixel (from 0 to volume radius). 0 = no cutoff

Resolution level [Å]

The target resolution level [in Angstrom]

Optional experimental data

Provides a reference 3d structure which is used instead of the spider x-ray curve for correction

3d volume

Provides the 3d volume to be sharpened

1d rot avg of power spec

1d curve showing the 1d averaged power spectrum of the input 3d volume

Enhancement curve

1d curve showing the 1d representation of the output “Amplitude filter”.

Amplitude filter

3d volume that shows the applied sharpening filter image.

Amplitude Corrected 3d volume

Contains the corrected output 3d volume.

Written Header Values
  • resolutionLevel Resolution where cut off was performed
  • pixelSize Pixel size of the volume

This logic is not computationally heavy but needs a lot of RAM for execution. The biggest tested dimensions were 1024x1024x1024 which occupied roughly 12gb of RAM. If not enough RAM is available, this logic will fail to execute.