[Chimera-users] Analysis of NMR-like ensembles in CHIMERA

[Elaine Meng]

Hi Gleb
Eric explained the numbers in a previous reply:
<http://plato.cgl.ucsf.edu/pipermail/chimera-users/2016-May/012304.html>

I would only clarify that although he said “percentage” it is really fraction”: without weighting, 0.5 means there is a contact or H-bond between the two residues in half (50%) of the frames used for the calculation. With weighting, it depends on the number of contacts or h-bonds between the pair, i.e. if there were 2 H-bonds between the pair of residues, it would count double.

It is also explained in the documentation, under the “Weight interactions…” option:
<http://www.rbvi.ucsf.edu/chimera/docs/ContributedSoftware/movie/movie.html#rins>

You would probably need to do some kind of scripting (sorting, string comparison) to make sure you are comparing the numbers for the same residue pairs, but I cannot advise on the details.
Best,
Elaine

On Jun 15, 2016, at 3:14 AM, James Starlight <jmsstarlight at gmail.com> wrote:

> Thanks for the suggestions again, Elaine!
> 
> For my particular case calculation of the contact pairs from the
> MDmovies produced reasonable results.
> Going to the analysis of the output logs
> 1) what determine the statistical value - at the end of each string e.g
> 
> ALA 252.B    vdw    GLY 75.C    1.4
> ALA 71.C    vdw    ILE 117.B    1
> ASP 178.B    vdw    GLY 54.C    1.2
> GLN 120.B    vdw    HIS 53.C    1.2
> ARG 198.B    vdw    GLU 64.C    0.6
> ALA 252.B    vdw    ASN 68.C    1.4
> 
> does the value of 1.4 corresponds to most likely contact (observed on
> bigger number of snapshots) than 0.6 for intance?
> 
> 2) Obtaining two such logs where the string order (contact pairs
> ranged to its statistical weights) are mixed within it- what are the
> trivial suggestion for the comparison such two logs to establish
> shared contact pairs with equal statistical weights?
> 
> Thanks!
> 
> Gleb
> 
> 2016-06-14 18:14 GMT+02:00 Elaine Meng <meng at cgl.ucsf.edu>:
>> Hi Gleb,
>> As you probably saw already, you can calculate a contact map for each ensemble.
>> <http://www.rbvi.ucsf.edu/chimera/docs/ContributedSoftware/movie/movie.html#rins>
>> 
>> However, there is nothing to do such a statistical analysis automatically.  For each contact map, you would just get a set of edges and their values.  Maybe if you had a large number of snapshots (frames) for each of the two ensembles, you could take several different samples from each one (thus multiple contact maps from each ensemble) and then somehow use all those edge values to calculate some statistics.  However, I am not a statistical expert and cannot advise on what methodology would be valid or most appropriate for your data.
>> 
>> Best,
>> Elaine
>> 
>>> On Jun 14, 2016, at 6:41 AM, James Starlight <jmsstarlight at gmail.com> wrote:
>>> 
>>> Thanks so much Elaine!
>>> 
>>> To simplify the problem -
>>> I have 2 cases both of which are the same process: binding established
>>> between A and B caputred by means of i) docking ii) MD simulation.
>>> 
>>> Let's assumes that contact map is the most important criterioum for
>>> my case - because one of the *docking* ensemble was driven by means of
>>> experiemtal restraits so the contacts between pairs of the residues of
>>> A and B can be considered here as the reference which I want to
>>> reproduce in the second MD ensemble in unbiassed fassion.
>>> 
>>> Is it possible to calculate contact maps for both ensembles separately
>>> using MD movie and than calculate some statistical variance or RMSD of
>>> established contact pairs between A and B within each of the ensembles
>>> and after all to compare both outputs to understand what are the
>>> cotacts from the MD ensembles has most statistical significance?
>>> 
>>> Gleb