Dear Community,
I'd like to continue making small additions to the RELION 1.4 code base through vlion. The Volta phase plate has been gaining traction in the community lately, so I think it's time to adapt the processing. The latest update (https://github.com/dtegunov/vlion) includes a CTF model with phase shift. There are two places you can tell vlion its value:
*_ctffind3.log: These files are generated by CTFFIND and used in 1.4 to obtain the CTF model for whole micrographs, e. g. for autopicking. While CTFFIND 4 offers support for phase shift fitting, I haven't found a way to obtain it through the legacy interface used by RELION to run it. Thus, the _ctffind3.log files must be constructed manually based on whatever output CTFFIND 4 provides. RELION grabs the values from the line ending with "Final Values". If that line includes 5 instead of 4 values, vlion will parse the 5th value as the phase shift. Maybe someone from the Grigorieff lab could write a conversion script? I'm using my own fitting tool, which will be released later this year.
*_data.star: Put the phase shift values into a "rlnPhaseShift" column. Note that RELION won't be able to read the file with this column included.
In both cases, phase shift values are given in degrees.
Furthermore, if you'd like to try some really exotic CTFs, there is now the option to read them from 2D images (or image stacks) referenced in the "rlnCtfImage" column in *_data.star. This was only implemented for 3D particles in a sub-tomogram averaging workflow previously. Other than for 3D, the data have to be in FFTW layout, i. e. center at 0, and only the non-redundant half. Activate this by running ml_optimiser with --use_custom_ctf.
I used this code to process a 20S proteasome data set acquired with a VPP out-of-focus. The final resolution was 3.6 A, based on ca. 11 000 particles with D7 symmetry. This was not the same data set as in the recent Danev & Baumeister paper (http://elifesciences.org/content/5/e13046v1), and it seems the sample quality was the limiting factor here.
Cheers,
Dimitry
|