The fit from CTF estimation using CTFFIND4 is much worse than when using CTFFIND3
Added by @Maria Janssen over 8 years ago
Hi!
I ran both CTFFIND3 and CTFFIND4 on a set of negatively stained images. On the SAME negatively stained image, both estimated the defocus values well, but CTFFIND3 gave a much better fit than CTFFIND4. See attached images. Any idea why?
Thanks!
Mandy
| CTFFIND4.png (575 KB) CTFFIND4.png | CTF image report from CTFFIND4 | ||
| CTFFIND3.png (647 KB) CTFFIND3.png | CTF image report from CTFFIND3 |
Replies (5)
RE: The fit from CTF estimation using CTFFIND4 is much worse than when using CTFFIND3 - Added by Anchi Cheng over 8 years ago
We only discovered that this week as well. ctffind3 and ctffind4 define their astigmatism angle differently but we did not know about it. See Issue #4047. After I test it, I will merge into 3.2
RE: The fit from CTF estimation using CTFFIND4 is much worse than when using CTFFIND3 - Added by Anchi Cheng over 8 years ago
Merged into 3.2 please update and give it a try.
RE: The fit from CTF estimation using CTFFIND4 is much worse than when using CTFFIND3 - Added by Michael Cianfrocco over 8 years ago
As a follow up to Maria's question (which I'm glad is resolved!), we have noticed that CTF confidence values are lower for images that are binned by 2 after frame alignment. Is this a bug or an observation that others have had for binning aligned, summed movie frames by 2?
Attached are images from no binnning (pixel size = 0.6 A/pix) or when the movies were binned by 2 (pixel size = 1.2 A/pix). Since the binned image has information to 2.4 Angstroms, I wouldn't expect that this would affect the CTF confidence fit.
What do you think?
Thank you!
Mike
| ctffind3_bin1.tiff (672 KB) ctffind3_bin1.tiff | CTFFIND3 - No binning after movie aligning and summing | ||
| ctffind3_bin2.tiff (603 KB) ctffind3_bin2.tiff | CTFFIND3 - Bin by 2 after movie aligning and summing |
RE: The fit from CTF estimation using CTFFIND4 is much worse than when using CTFFIND3 - Added by Michael Cianfrocco over 8 years ago
As a follow up question, we have also always noticed that the fitted line and experimental line for the 'Correlation' plot always are a little bit different from each other, starting at low resolution. It seems that other plots (like the one Mandy showed above) don't have this feature. Is this something we should look into?
RE: The fit from CTF estimation using CTFFIND4 is much worse than when using CTFFIND3 - Added by Anchi Cheng over 8 years ago
I don't have answers for these. The offset might be related to amplitude contrast ?
I started an issue for these #4241 and asked Neil Voss who wrote the CTF validation part to take a look. Please attach a pair of example mrc images there along with pixel size, Cs, etc. If they are too big, contact him directly and arrange some dropbox solution.
Also. please specify the version of Appion you ran into this problem to help with diagnosis.