GROMACS version: 2019
GROMACS modification: No
Hello,
I am woundering if there are some possibilities to define selection for gmx cluster? For example I need to perform clustering of snapshots based on the centre of mass (COM) of the particular residue (ligand)? Can I use -select option with the expression like “resname LIG and COM” for this purpose?
Thank you in advance!
Hi,
currently, gmx cluster does not support the option -select. You can provide an index file using the option -n
Kind regards
Alessandra
Thank you Alessandra!
Is it possible to create selection (e.g. within index file) of the defined residue and its COM using some of gromacs Utility besides of make_ndx ? It could be useful to automatize the workflow supposing that I execute some shell script in loop for the ananysis of several trajectories. I would be grateful for some example if it exists.
Cheers,
One extra question - is it possible in addition to obtain clustering histograms (not a rmsd graph) e.g. showing number of conformations in separate clusters depending on its energy or other criteriums ?
Hi,
to generate index file you can also use
gmx select (http://manual.gromacs.org/current/onlinehelp/gmx-select.html?highlight=gmx%20select )
Kind regards
Alessandra
Hi,
gmx cluster can provide with option -sz, an xvg file with the cluster sizes. The available method for cluster determination are listed in the help of gmx cluster (http://manual.gromacs.org/current/onlinehelp/gmx-cluster.html?highlight=gmx%20cluster).
gmx cluster covers a selection of methods, more methods are available in literature.
Kind regards
Alessandra
Thank you Alessandra!
It works fine!
I am going to create another topic for gms select so …
one extra question regarding cluster visualisation - would be possible additionally to sort the computed clusters based on either lowest energy or average energy? What option for gmx cluster could be useful?
Hi,
There is no specific option for this in gmx cluster.
In general, one can get the difference in free energy between 2 states of the system from the ratio of the population of the states (statistical mechanics), assuming that you have sufficient sampling of phase space. gmx cluster provide the clusters’ population, from the population you can get the free energy difference.
Kind regards
Alessandra
1 Like
Thank you Alexandra!
You are righy, but my task alitle bit different - sorry I did not specify it correcly. Actually I am dealing with the cluster analysis of ligand conformations obtained from docking run. This is typical multi-model pdb file which has 20 models of the ligand, containing also Free energy of binding (in the REMARK of each model). I’ve already clustered this ensemble based on the RMSD using gmx cluster. But now I would like also to add the energy value for each of the conformer to my cluster diagramm in order that I could see how the clusters are arranged according to the energy of each conformers.
Hi James,
There is no specific gmx tool, doing what you ask.
One option can be to plot the RMSD values against free energy values (using one model as reference). On the top of the plot you can locate the central structure of gmx cluster analysis. This can allow you to see the relation between the gmx cluster analysis and the free energy energy. But this is just a general personal suggestion.
Best regards
Alessandra
1 Like
thank you very much Alessandra for this very useful suggestions!
In fact, someone has suggested me to add the energies (e.g. stored in the separate clusters) to the results of gmx cluster but I could not either understand what they were talking about.
One additional question: If I would like to change the RMSD-based clusterization to the RMS distance (determine clusters based on the distances of the structures, rather than by its internal similarities), should I only add -dista flag or additionally I should to switch to “single linkage” method of clusterization ?
Hi,
In gmx cluster RMS deviation after fitting or RMS deviation of atom-pair distances can be used to define the distance between structures.
You can choose to use RMSD of distances instead of RMS deviation using the option -dista.
Is this what you want?
Best regards
Alessandra
OK, I got the idea, thank you again Alessandra!
I suppose that I hava already achived what I was looking for via option -nofit (meaning that I don’t compare internal structures of the conformations but rather cluster them based on the deviation of the distances between them). Aditionally I’ve tried also to add -dista but in that case all of the conformations fell into the same cluster