[gmx-users] Clustering based on other parameters than RMSD

[soumadwip ghosh]

Hi all,

Sorry about the last incomplete mail.I have a trajectory of 300 ns for a
membrane protein. I want to sample native/non-native conformations and
compare across various systems. My question is can parameters like
number/occupancy of hbonds or the time evolution of helical residues
(output from gmx do_dssp) be clustered using something like gmx cluster or
something similar? As far as I understand gmx cluster does RMSD/distance
RMSD clustering based on a predefined cut-off. I am trying to extract the
number of accessible microstates for a given system, extract the
corresponding trajectories and do some property analysis. This is to be
done based on proportion of native/non-native like states which may be
related to assigning partially denaturated state of the protein. I did
cluster analysis on PCA but it seems difficult to compare across different
systems since they have different principal components. I have plotted free
energy surfaces by taking the percentage helicity, number of hydrogen bonds
and their probability (using the gmx sham utility) as X, Y and Z variables
respectively and I am looking at different microstates. Is this approach
right or am I missing something?

I would appreciate any input on this regard.

Thanks and regards,
Soumadwip Ghosh
Post Doctoral Research Associate
Prof. Vaidehi Nagarajan's group
Department of Molecular Immunology
City of Hope Cancer Research Center
Duarte, CA 91010
United States