[3dem] [TEM] CTF and SNR

[John Rubinstein]

Hi Steve,

As Greg McMullan and Richard Henderson showed recently, amorphous ice does give Thon rings. However, ice also exhibits a beam induced ‘Brownian’-like movement so that the power spectrum from the average of DDD frames does not often show Thon rings.
https://www.ncbi.nlm.nih.gov/pubmed/26103047 <https://www.ncbi.nlm.nih.gov/pubmed/26103047>
Alexis Rohou modified CTF find in order to allow use of this signal in CTF estimation.

Best regards,
John



-- 
John Rubinstein
Molecular Medicine Program
The Hospital for Sick Children Research Institute
686 Bay Street, Rm. 20-9705
Toronto, ON
Canada
M5G 0A4
Tel: (+001) 416-813-7255
Fax: (+001) 416-813-5022
www.sickkids.ca/research/rubinstein

> On Jun 7, 2017, at 9:33 AM, Ludtke, Steven J <sludtke at bcm.edu> wrote:
> 
> Hi Philip,
> pure water ice, alas, does not exhibit Thon rings (at least not significantly). Other buffer constituents may. This makes modeling noise a bit tricky. 
> 
> Many of the methods in the field start with the assumption that the image consists of a true projection of a macromolecule, convolved with the CTF with added uncorrelated Gaussian noise. That is, noise with a flat power spectrum and no correlation among nearby pixels. Prior to direct detectors and counting, the background noise power spectrum in typical cryoEM images had an exponential decay, and could with a little processing be "flattened" to produce something more or less like uncorrelated Gaussian noise.  The problem with this assumption, even then, is that CryoEM images actually contain a lot of other stuff, and from the perspective of single particle reconstruction, noise isn't just the scattering off the solvent, it's anything present in the image which is not a perfect projection of the particle. 
> 
> This is what led to EMAN2's approach of computing the power spectrum of masked out particles, and masked out regions outside particles to compute an accurate estimate of the SNR of the particles within the images. Whether having this really produces better reconstructions is a separate question, but making this estimate DOES tell you quite a lot about particle data quality and whether you should expect to be able to achieve high resolution from your data.
> 
> I'm attaching some slides from about 10 years ago showing how SSNR estimates assuming a pretty flat background can give completely inaccurate estimates of the actual SSNR of the images. Hope they are reasonably self-explanatory. This compares the approach of considering noise to be anything in the power spectrum without Thon rings to be noise, with the approach of computing the noise from regions of the image not containing particles.
> 
> https://drive.google.com/file/d/0BxreryX3Fp-ZUVJtTFlYX0lmbzg/view?usp=sharing <https://drive.google.com/file/d/0BxreryX3Fp-ZUVJtTFlYX0lmbzg/view?usp=sharing>
> 
> ----------------------------------------------------------------------------
> Steven Ludtke, Ph.D.
> Professor, Dept of Biochemistry and Mol. Biol.         (www.bcm.edu/biochem <http://www.bcm.edu/biochem>)
> Co-Director National Center For Macromolecular Imaging        (ncmi.bcm.edu <http://ncmi.bcm.edu/>)
> Co-Director CIBR Center                          (www.bcm.edu/research/cibr <http://www.bcm.edu/research/cibr>)
> Baylor College of Medicine                             
> sludtke at bcm.edu <mailto:sludtke at bcm.edu>
> 
> 
> 
> _______________________________________________
> 3dem mailing list
> 3dem at ncmir.ucsd.edu
> https://mail.ncmir.ucsd.edu/mailman/listinfo/3dem

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.ncmir.ucsd.edu/pipermail/3dem/attachments/20170607/d6aff563/attachment.html>