Hi everyone,
Thank you all for your helpful suggestions.
It turned out that inverting a mask with e2proc3d resulted in a volume with values from -1 to 0. The volume still looks fine in Chimera but Relion expects values from 0 to 1. Adding +1
to each voxel using relion_image_handler fixed this issue and signal subtraction is now working.
Best,
Jan
Reza Khayat, PhD
Assistant Professor
Department of Chemistry
City College of New York
85 Saint Nicholas Terrace, CDI 2.318
New York, NY 10031
212-650-6070
From: Collaborative
Computational Project in Electron cryo-Microscopy [mailto:[log in to unmask]] On Behalf Of Craig
Yoshioka
Sent: Monday, May 07, 2018 2:25 PM
To: [log in to unmask]
Subject: Re: [ccpem] Help with signal subtraction in Relion
I’m interested in what other have to say as well.
I’ve also seen this a couple of times. I’ve gotten signal subtraction to work “OK” for this use case, but I think it’s particularly messy for two reasons:
(1) the micelle/nanodisc is fairly heterogenous and subtracting an average leaves residuals in the raw data that make it difficult to completely remove the signal. The residuals also make it more challenging to improve the alignment, which is
usually the point of trying this. Signal subtraction (IMO) strongly favors the use case of first subtracting the best aligned and well-ordered part of the map to help improve the remainder.
(2) I also suspect that signal subtraction in general is very reliant on accurate background estimation and scaling. I could be wrong, but the current approaches seem do a good enough job from the POV of averaging all the data together, but
not necessarily for getting accurate subtractions.
In one of our projects we are working with a membrane protein reconstituted into nanodiscs. We now wanted to test if subtracting the nanodisc signal can give us a better
result compared to using a mask excluding the nanodisc density applied after the first conversion in 3D refinement. In all attempts at subtraction so far, the nanosdic was still present in the resulting maps and we were wondering if this is normal or if we
are missing something in the process. Sure, the density disappears when playing with the threshold in Chimera, but this was already the case before signal subtraction.
Here is our current workflow for signal subtraction:
-
Generate a soft mask around the protein using Relion (v 2.0.3)
-
Invert the mask using e2proc3d with the argument --mult=-1
-
Run signal subtraction in Relion using the angular assignments and model of 3D refinement
-
Rerun 2D classification, 3D classification or 3D refinement using the resulting subtracted particles
I made sure that both map and mask have the same box size, and that both are aligned. In case we are subtracting everything but one monomer we can remove the other subunits,
but this only shows when playing with the threshold in Chimera. The mask definitely small enough to exclude the nanodisc.
Any help or suggestion would be highly appreciated.
