Fwd: [External] Boost value in aMD simulation

From: Thomas Evangelidis (tevang3_at_gmail.com)
Date: Tue Aug 19 2014 - 05:24:17 CDT

On 19 August 2014 12:56, James Starlight <jmsstarlight_at_gmail.com> wrote:

> Thomas,
>
> I've noticed that in case of *long* loops the big uncertainly exist during
> its clustering using only rmsd criterium: for instance with the big cutoff
> (e.g 3.5-4 A)all
>

This is a big cutoff.

> conformations could be grouped in the few (up to 10) clusters with small
> numbers of structures not in one of the clusters (outliers?) but in this
> case representative structure from every cluster are not very differs from
> each others in conformations (because most of its part has not any 3D
> structure). BTW do you know any software
>

Do you mean secondary structure here?

> which could produce distribution patterns on the diagram grouping all
> clustered together (I'm using vmd cluster plugin where I could not find
> such representations)?
>
>
Like a dendrogram? I usually do this in R but using it's internal
clustering algorithms.

However, my impression is that you spend too much time in the preliminaries
of the project. The essence is the simulation with the ligands. So just
pick up one starting conformation and move forward. You may spend many
weeks trying different clustering schemes and still not find a solution
that satisfies you.

2014-08-18 11:21 GMT+02:00 Thomas Evangelidis <tevang3_at_gmail.com>:

The right way is to reweight the clusters, but as I stated before
> reweighting aMD is not accurate for such big systems. If the number of
> frames is not too big and you can do clustering, just go for that solution
> which is simpler.
>
>
>
> On 18 August 2014 11:25, James Starlight <jmsstarlight_at_gmail.com> wrote:
>
>> Some additional question about to test convergence of my trajectories in
>> case of loop refirement:
>>
>> e.g I have 10 trajectories calculated for the same system initiated from
>> different started velocities and with slightly different aMD boost values.
>> What might be better:
>>
>> 1) Merge all trajectories into 1 big trajectory and perform i) cluster
>> analysis to determine shared clusters in the selected refined (loop)
>> region= > test onto the structural convergence ii) PCA of this big
>> trajectory => determine shared modes of the collective motions seen in all
>> 10 trajectories => test onto the dynamical convergence
>>
>> Might it be useful?
>>
>> James
>>
>>
>> 2014-07-24 14:13 GMT+04:00 Thomas Evangelidis <tevang3_at_gmail.com>:
>>
>>
>>>
>>>
>>> On 24 July 2014 12:54, James Starlight <jmsstarlight_at_gmail.com> wrote:
>>>
>>>> Thanks again, Thomas!
>>>>
>>>> I guess that the combination of steered and accelerated MD could be
>>>> very effective method to monitor pathway of the ligand motion from the
>>>> water exteriour to the receptor orthosteric-binding pocket interiour (what
>>>> is exactly my goal at this moment!).
>>>>
>>>> Some techniqual questions:
>>>>
>>>> 1) could the averaged structure from the largest cluster be assumed as
>>>> the representative structure?
>>>>
>>>>
>>> I didn't mention the word "averaged", I mention the cluster
>>> representative. The answer is no. It may not be native like at all
>>> -especially for a loop of 30 aa-, but it's the closest you can get with the
>>> current computational tools.
>>>
>>>
>>>> 2) I guess that processing of 1 trajectory made from seveal tens of
>>>> pdbs by means of some md tools (e.g I coould look at amber processing
>>>> besides the gromacs tools) is good sollution. Do you know some software for
>>>> conversion of multi pdb to amber-like or gromacs-like trajectories ? In
>>>> past I've made only NMR-like format pdb from multi single pdbs.
>>>>
>>>>
>>> VMD.
>>>
>>>
>>>> Best,
>>>>
>>>> James
>>>>
>>>>
>>>>
>>>>
>>>> 2014-07-24 13:05 GMT+04:00 Thomas Evangelidis <tevang3_at_gmail.com>:
>>>>
>>>>
>>>>>
>>>>>
>>>>> On 24 July 2014 09:13, James Starlight <jmsstarlight_at_gmail.com> wrote:
>>>>>
>>>>>> Thomas, many thanks for the suggestion!
>>>>>>
>>>>>> So Rosetta in comparison to modeller could be used for i) prediction
>>>>>> of ss eleents in *long* loops ii) performing clustering of generated models
>>>>>> based on chosen criterium (e.g conformation of loop, or % of SS in its),
>>>>>> couldn't it?
>>>>>>
>>>>>> Bare in mind that you can restrain the secondary structure of the
>>>>> vestibule to helix with both software. As for clustering, I usually convert
>>>>> the models (.pdb files) to a trajectory and do the clustering with normal
>>>>> MD analysis tools (e.g. g_cluster). But there is also a stand-alone
>>>>> application named calibur that can read and cluster the pdb files directly.
>>>>>
>>>>>
>>>>>
>>>>>> Regarding the general workflow of the modelling and refirement based
>>>>>> on your assumptions I guess it would be more correct to do:
>>>>>>
>>>>>> 1) homology modelling by modeller/posetta-> clustering -> selection
>>>>>> of model from bigeest cluster
>>>>>>
>>>>>> selection of the representative structure of the largest cluster.
>>>>>
>>>>>
>>>>>> 2) short cmd simulation in membrane to relax the system
>>>>>>
>>>>>>
>>>>> By all means!
>>>>>
>>>>>
>>>>>> 3)loop refirement w/o membrane with applied position restraints (yes,
>>>>>> I dont like to kill this idea :-) )
>>>>>>
>>>>>>
>>>>> I doubt about the necessity of this step. I would proceed directly to
>>>>> production run, although I have already stated my objections about your
>>>>> workplan. If binding to the main ligand pocket is your ultimate goal, then
>>>>> you should steer the ligand to that direction. In contrast, if your goal is
>>>>> detection of allosteric pockets then I doubt that aMD is the right enhanced
>>>>> sampling method.
>>>>>
>>>>>
>>>>>
>>>>>> James
>>>>>>
>>>>>>
>>>>>
>>>>> --
>>>>>
>>>>> ======================================================================
>>>>>
>>>>> Thomas Evangelidis
>>>>>
>>>>> PhD student
>>>>> University of Athens
>>>>> Faculty of Pharmacy
>>>>> Department of Pharmaceutical Chemistry
>>>>> Panepistimioupoli-Zografou
>>>>> 157 71 Athens
>>>>> GREECE
>>>>>
>>>>> email: tevang_at_pharm.uoa.gr
>>>>>
>>>>> tevang3_at_gmail.com
>>>>>
>>>>>
>>>>> website: https://sites.google.com/site/thomasevangelidishomepage/
>>>>>
>>>>>
>>>>>
>>>>
>>>
>>>
>>> --
>>>
>>> ======================================================================
>>>
>>> Thomas Evangelidis
>>>
>>> PhD student
>>> University of Athens
>>> Faculty of Pharmacy
>>> Department of Pharmaceutical Chemistry
>>> Panepistimioupoli-Zografou
>>> 157 71 Athens
>>> GREECE
>>>
>>> email: tevang_at_pharm.uoa.gr
>>>
>>> tevang3_at_gmail.com
>>>
>>>
>>> website: https://sites.google.com/site/thomasevangelidishomepage/
>>>
>>>
>>>
>>
>
>
> --
>
> ======================================================================
>
> Thomas Evangelidis
>
> PhD student
> University of Athens
> Faculty of Pharmacy
> Department of Pharmaceutical Chemistry
> Panepistimioupoli-Zografou
> 157 71 Athens
> GREECE
>
> email: tevang_at_pharm.uoa.gr
>
> tevang3_at_gmail.com
>
>
> website: https://sites.google.com/site/thomasevangelidishomepage/
>
>
>

-- 
======================================================================
Thomas Evangelidis
PhD student
University of Athens
Faculty of Pharmacy
Department of Pharmaceutical Chemistry
Panepistimioupoli-Zografou
157 71 Athens
GREECE
email: tevang_at_pharm.uoa.gr
          tevang3_at_gmail.com
website: https://sites.google.com/site/thomasevangelidishomepage/

This archive was generated by hypermail 2.1.6 : Wed Dec 31 2014 - 23:22:44 CST