Srdjan Masirevic

Chimera does exactly what you asking for. Download and install Chimera software:

https://www.cgl.ucsf.edu/chimera/download.html

Once it's installed, open your .pdb file containing the NMR conformers and go:

Tool/Structure Comparison/Ensemble Cluster

Then, a window will pop up and press "Apply". That way you will cluster an ensemble of NMR-derived protein structures into conformationally related subfamilies conformers...

see more information here:

https://www.cgl.ucsf.edu/chimera/docs/ContributedSoftware/ensemblecluster/ensemblecluster.html

hope this helps!

Dear Mario,

thanks for your answer it was very helpful. I actually use Chimera for visualization and structure editing but never came across clustering.

Do you know if I can set up specific number of clusters, for example out of 20 conformers I want three clusters so that I can pick three representatives?

Best,

Srdjan

Srdjan Masirevic

I would recommend check for the literature first and then decide if is recommendable setting up a defined number of clusters ad hoc...

https://academic.oup.com/peds/article/9/11/1063/1590384

after reading the literature, if you are still, interested in having 3 clusters only, I will post later on a method (with much more steps and more complicated than chimera's one) involving clustering using the k-means method where you can define ad hoc how many clusters do you want... I just need to check if the results are similar to those of obtained with chimera and I'll be back...

cheers!

Dear Mario,

I am comparing to some extent Xray to NMR so I am clustering it just because to get more NMR structures so that comparison can be head to head. Therefore, if I have 5 X-ray structures of the same protein I would like to have 5 clusters or 5 conformers, each conformer from each cluster. In some cases I might have 3 X-ray structures, therefore I would need 3 clusters etc, thats the only thing. Also would be great if can get the way to grep the most structurally diverse conformers from one NMR ensemble.

Any kind of help would work in this regard!

Many thanks!

Best,

Srdjan

Hey Mario,

I have found the way with Schrodinger to do structure ensemble NMR clustering with k-medoids.

I think that would work the best cus with k-medoids I get conformations that differ most from each other plus I can assign different number of clusters as a result and therefore change the cutoff of RMSD value.

Many thanks once again!

Best,

Srdjan

Hi Srdjan Masirevic

I will say that picking up three conformers that are most different from each other over 20 conformations is make no sense to anyone.

You will have to try over one thousand combinations to find out the three (C20, 3).

We are used to derived RMSD values with respect to "a reference structure" which could be a minimum free energy structure based on MD simulation or other structures that have been determined. If you wanna do this, I will suggest you use PyMol.

You can simply use PyMol and run the pymol script in the attached file. Please follow up with the step-by-step instruction:

1. Download this script in a certain directory

2. Use PyMol to open 20 structures and a reference structure

3. type commands in PyMol command line:

run /download_directory/align_all.py

align_all target=your_reference_structure

You will get the alignment of all models in a list to one target.

Best,

Yun-Tzai

Dear Yun-Tzai,

Thank you for your asnwer. I agree if I want to pick up three most diverse one, it will take number of calculation but with a good script it takes me 3 seconds.

I haven't express myself correctly. I have already made use of first one, energy most minimized conformation and now I am trying to add three more that are most diverse. Therefore clustering them into 4 clusters by applying k-medoid clustering helps me solve this problem. To make things simpler I have applied clustering only for the binding pockets of each conformer (10 Å distance from the ligand). You said that taking few conformers that are structurally different from each other makes no sense for anyone but for what I want to achieve it works for me.

The method you proposed doesn't really help me that much cus I have a bit simpler one that helps me achieve my goal more precisely.

Best,

Srdjan

Please read the following article:

Efficient molecular docking of NMR structures: Application to HIV-1 protease

Sheng-You Huang and Xiaoqin Zou

Protein Sci. 2007 Jan; 16(1): 43–51.

To find the 3 most diverse,calculate all 190 RMSD and sort the values to fish the

3 biggest ones.If you want to cluster them and know how many clusters you expect use a standard K-means algorithm (e.g R-package for statistics) probably

implemented in Chimera ,Maestro etc.

Mario E. Valdés-Tresanco, currently i am using Chimera for clustering but how can i know the cutoff used by chimera? i searched for this but i found that it is done automatically. is there a way to know what cutoff used after clustering?

Ibrahim M. Ibrahim if you are using Ensemble Cluster from chimera, you can check the help online for this module and also this publication...

An automated approach for clustering an ensemble of NMR-derived protein structures into conformationally related subfamilies. Kelley LA, Gardner SP, Sutcliffe MJ. Protein Eng. 1996 Nov;9(11):1063-5

hth!