Dear Eric, dear all,

 

if you want to do local MSA, focused 2D or 3D classification and/or focused refinement, the region to be analysed should be significantly larger than the specific area of interest. This is needed to catch the associated conformational changes in the proximity of the area of interest. If your mask comprises only the drug/inhibitor etc. the classification or refinement will become unstable. Increasing the mask area by ~20/30/40 Å can help (depending on object size, size of region of interest etc.). It is also worth trying both multivariate statistical analysis and maximum likelihood approaches if one or the other doesn’t work. MSA seems more sensitive to smaller changes.

For some original concepts on local / focused 2D and 3D analysis and recent reviews & applications you may want to have a look at some of these papers:

 

https://www.ncbi.nlm.nih.gov/pubmed/14985767       concept of local MSA, usage of mask for focused classification

https://www.ncbi.nlm.nih.gov/pubmed/18758445       3D structure sorting based on resampling & 3D classification

https://www.scirp.org/html/17-1240596_62411.htm  (and references therein)   local MSA with mask, resampling & 3D classification

https://www.ncbi.nlm.nih.gov/pubmed/28850874       review on focused classification and refinement, including aspects on mask size

https://www.ncbi.nlm.nih.gov/pubmed/27730650       review including aspects on structure sorting (3 approches to it: (i) template-based methods, (ii) classification based on statistical analysis using MSA and bootstrapping methods and (iii) ML-based sorting), see references therein

https://www.ncbi.nlm.nih.gov/pubmed/29143818       an example where focused refinement was essential to help improving the map (there are many more in the field)

 

Best,

 

Bruno

 

 

###########################################################################

Bruno P. Klaholz

Centre for Integrative Biology

Department of Integrated Structural Biology
Institute of Genetics and of Molecular and Cellular Biology
IGBMC

http://igbmc.fr/Klaholz

http://www.igbmc.fr/grandesstructures/cbi

http://frisbi.eu

http://instruct-eric.eu

 

 

 

 

 

From: Collaborative Computational Project in Electron cryo-Microscopy [mailto:[log in to unmask]] On Behalf Of Eric Hanssen
Sent: 13 July 2019 01:29
To: [log in to unmask]
Subject: Re: [ccpem] [3dem] 3D classification

 

Thank you all for the suggestions. I ll give it a try see how that goes ...

Cheers

Eric

 



 

----------------------------------------------------------------------------

Assoc Prof Eric Hanssen

Principal Research Fellow

Head Bio21 Advanced Microscopy Facility

Bio21 Molecular Science and Biotechnology Institute

30 Flemington Road - The University of Melbourne - Victoria 3010 - Australia

email: [log in to unmask] | Office: +61 3 83442449 | Microscope: +61 3 83442509 | Fax: +61 3 9347 826

 


On 13 Jul 2019, at 03:13, Penczek, Pawel A <[log in to unmask]> wrote:

Hi

 

there has to be some kind of misunderstanding here as in general it is impossible to perform eigeanalysis in the way described in the manual. 

Regards,

Pawel Pawel


On Jul 12, 2019, at 11:53 AM, Ali Punjani <[log in to unmask]> wrote:

**** EXTERNAL EMAIL ****

Hi Eric,

As a direct method to resolve the heterogeneity, you may wish to try the new 3D Variability Analysis method available in cryoSPARC v2.9 (released earlier this month). This algorithm can detect very small conformational changes at high resolution, by computing the eigenvectors of the 3D density covariance of the density. These eigenvectors correspond to the modes of maximal variability in the particle data, and therefore can be used to resolve both discrete and continuous heterogeneity. More details, including examples and a detailed tutorial can be found here:

3D Variability Analysis can also be used with a mask, to focus on the relevant region of the structure, ignoring conformational variability that may be occurring in other irrelevant regions (e.g. it can help to mask out the micelle or nanodisc for a membrane protein).

Note that, as Sjors already mentioned, the major determinant of success will be detectability of the change caused by binding. If there is no conformational change associated with the binding, it is unlikely that the ligand alone would cause enough change in each particle image to be statistically distinguishable from noise, regardless of the computational technique used.

 

Good luck with your processing!

Ali


It can also be used in a masked fashion,

On Thu, Jul 11, 2019 at 8:24 PM Eric Hanssen <[log in to unmask]> wrote:

Hi all,

Before I commit some computer resource to a job I’d rather know if it is doable.

 

I have 200,000 particle of a 600kDa protein, a current map at 3.4A that looks very nice, I do not need a better resolution for now. I have a drug (dipeptide) attached to the protein but  only a subset of the protein has the drug bound. Is it possible with a 3D classification to separate such a small difference? I don’t mind which software, relion, cryosparc, eman2 ….

 

Or is there a better way, eg. 3D classification of a very small box size centered around the binding site?

 

Cheers

Eric

 

 

-----------------

Assoc. Prof Eric Hanssen

 

Head - Advanced Microscopy Facility

Honorary Principal Fellow – Department of Biochemistry and Molecular Biology

 

President Australian Microscopy and Microanalysis Society


Bio21 Molecular Science and Biotechnology Institute
30 Flemington Road - The University of Melbourne - Victoria 3010 - Australia
email: [log in to unmask] | Office: +61 3 83442449 | Microscope: +61 3 83442509

Web: www.microscopy.unimelb.edu.au

 

bio21   amms-logo

 

 

 

_______________________________________________
3dem mailing list
[log in to unmask]
https://mail.ncmir.ucsd.edu/mailman/listinfo/3dem

 

To unsubscribe from the CCPEM list, click the following link:
https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=CCPEM&A=1


To unsubscribe from the CCPEM list, click the following link:
https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=CCPEM&A=1