Implementing gmx clustsize with MDAnalysis

[Luis Itzá Vázquez-Salazar]

Dear Community of MDAnalysis,

I am working with some CHARMM trajectories and I would like to do a cluster size analysis similar to what it is done with gmx clustsize but using mdalysis. Could you assist me with some suggestions on how to start?

In advance, thank you very much!

Cheers,

Luis

[Oliver Beckstein]

Hi Luis,

Welcome to the MDAnalysis list!

On Aug 19, 2021, at 4:58 AM, Luis Itzá Vázquez-Salazar <litzav...@gmail.com> wrote:

Dear Community of MDAnalysis,

I am working with some CHARMM trajectories and I would like to do a cluster size analysis similar to what it is done with gmx clustsize but using mdalysis.

What does gmx clustsize do? How does it compute distance between frames, what clustering algorithm does it use?

Could you assist me with some suggestions on how to start?

The encore module https://docs.mdanalysis.org/2.0.0-dev0/documentation_pages/analysis/encore.html probably has some clustering functionality built in ; the User Guide has some starting information https://userguide.mdanalysis.org/stable/examples/analysis/trajectory_similarity/README.html 

You can also build the frame-vs-frame RMSD distance matrix yourself (using https://docs.mdanalysis.org/2.0.0-dev0/documentation_pages/analysis/rms.html#MDAnalysis.analysis.rms.RMSD with different reference frames) and then use a clustering algorithm from scikit-learn. 

In advance, thank you very much!

Sorry, I don’t have a ready-made solution for you. Maybe other users/developers can share what they’ve been doing.

Oliver

Cheers,

Luis

--

Oliver Beckstein (he/his/him)

email: orbe...@mdanalysis.org
twitter: @orbeckst

MDAnalysis – a NumFOCUS fiscally sponsored project

https://www.mdanalysis.org/

[Pablo G Garay]

Hi Luis,

Maybe this conversation in the group could be a first approximation to obtain the clusters size https://groups.google.com/g/mdnalysis-discussion/c/oBIBayv7vrY

Cheers,

Pablo