AR-DECON – Deconvolution to restore cryo-EM maps with anisotropic resolution.
AR-DECON uses deconvolution to compensate for missing information in Fourier space. Running AR-DECON on COSMIC2 will generate a deconvolved output map. If you want to sharpen further/modify the map, you will need to do this next.
An example of what AR-DECON does with a published map of kinesin-binding protein (EMDB 24677):
AR-DECON improved resolvability for this reconstruction. The original map had a sphericity of 0.85 from the 3DFSC server. You can also see that DeepEMhancer does not work as well. Importantly, AR-DECON is more transparent regarding how it modifies the reconstruction compared to the generative AI tool DeepEMhancer.
This is the original figure from our publication showing the Euler angle distribution:
And this is the output FSC curves from AR-DECON and 3DFSC:
Inputs:
- Input: Unsharped 3D reconstruction
- Do NOT use post-processed or masked maps
- Required: Half maps
- Optional:
- Mask – Apply the specified mask file during dFSC calculation and deconvolution.
- Sigma – Sigma value for generating OTF (default: auto-calculated).
- Smooth – Smoothing parameter for deconvolution (default: 0.5).
- Nonlinearity – Parameter for deconvolution (default: 10000).
- Iterations – Number of deconvolution cycles (default: 50)
Outputs
- The following files and extensions are output from AR-DECON:
- _Decon.log – Log file
- _Decon.mrc – Deconvoluted 3D reconstruction (not sharpened)
- _HalfMapdFSC1.png/svg/txt – Image files showing directional FSC curves
- _HalfMapdFSC3d.mrc – 3D representation of Fourier space
- _HalfMapdFSC3d.otf – optical transfer function
- _HalfMapdFSCAvg.txt – average FSC value
- _HalfMapFibonacciPoints.txt – sampling points for directional FSCs
- _Masked.mrc – Masked input (if mask provided)
Reference
Deconvolution to restore cryo-EM maps with anisotropic resolution.
Junrui Li, Yifei Chen, Shawn Zheng, Angus McDonald, John W. Sedat, David A. Agard, Yifan Cheng
bioRxiv 2025.02.23.639707; doi: https://doi.org/10.1101/2025.02.23.639707