search

ml-struct-bio / cryodrgn

Neural networks for cryo-EM reconstruction
http://cryodrgn.cs.princeton.edu
GNU General Public License v3.0
288 stars 74 forks source link

CTF mismatch with RELION #381

Open ma-gilles opened 3 weeks ago

ma-gilles commented 3 weeks ago

Hi,

I noted a small difference between cryoDRGN and RELION in how the wavelength is defined, which made a change of about 0.6% in the CTF when I tested it.

Here is the cryDRGN one: https://github.com/ml-struct-bio/cryodrgn/blob/2e9c376aee84ff354c2515a7a40d1e4128f48218/cryodrgn/ctf.py#L95

Here is the RELION one: lambda=12.2643247 / sqrt(local_kV (1. + local_kV 0.978466e-6));

// See http://en.wikipedia.org/wiki/Electron_diffraction

I am not sure how big the impact is, but just FYI!

Marc

zhonge commented 2 weeks ago

Thanks for reporting this Marc. This looks like a difference in sigfigs. @michal-g and I discussed, and our current plan is to make this change in the next major version of cryodrgn (4.x).