Re: Accelerated MD in NAMD

From: James Starlight (jmsstarlight_at_gmail.com)
Date: Thu Nov 07 2013 - 01:09:45 CST

Professor Wereszczynski,

also I've seen the implementation of the aMD to the GPCR system (work of
I.Tikhonova) where only dihedral boost have been applied to the protein and
lipids as two separate terms.
Does it means that the transition to the active state of the receptor is
primarily structural dependent event? (this time you have not used total E
boost which are influence on the diffusion rates). Doest it takes from the
assumption that the receptor is fixed in the membrane so we might not take
into account diffusion ?

some off-topic- in this paper I've seen interesting methodology when you
compare MD observations with the distribution of the X-ray structures of
this receptor (solved in active and inactive states as such reference
points). My question is in methodology- will it more correct to project
C-alpha coordinates of X-ray structures onto the Principal modes calculated
for the MD trajectory or alternatively project MD snapshots onto the
Principal modes calculated from the ensemble of the X-ray structure ?(!) In
last case we have robust assumption that X-ray ensemble is the trajectory
capturing R->R* transition seen in experiment (!) So lowest frequency modes
gives evidence about possible path of this transition if we have number of
intermediates. Alternatively in the first case (as has been done in the
Tikhonova paper) we have assumption that accelerated simulation with the
artificial boost can be the reference for the monitoring some biological
event (R->R* transition) so in this case we test X-ray structures (not the
simulation setup). Which statement would be most probably? Does somebody
see the implementation of such methodology (projections of the X-ray
structures onto MD and vice versa) with other proteins ?

James

2013/11/7 James Starlight <jmsstarlight_at_gmail.com>

> Thomas, Jeff, thanks for suggestions!
>
>
> I would clarify that
> 1) using second boost I've measured <Utotal> not Etotal. I suppose that it
> might be wrong due to need of including kinetic energy as well if we are
> dealing with the diffusion.
> So In my case the total energy would be bigger (than -360 which is
> potential only) and E<threshold> should be also bigger
>
> 2) Could the projection onto the lowest principal components (calculated
> from the accelerated trajectory)
> give some evidence about preservation (or not) the Boltzmann sampling of
> my system. In what cases I can conclude about distortion of such sampling
> (e.g by means of applying low alpha) looking only on that projections ?
>
> James
>
>
> 2013/11/7 Jeff Wereszczynski <jwereszc_at_iit.edu>
>
>> Thomas makes a good point, you should surely search the literature for
>> caveats about aMD applied to membrane systems. My experience is based
>> largely upon solvated protein systems, and as Yi's paper points out there
>> may be some undesirable effects from aMD on lipids.
>>
>> As for reweighting, the answer to that depends on who you ask :) IMHO,
>> reweighting aMD is generally a fruitless endeavor, the differences in boost
>> potentials you get in a simulation are way too large to use precise
>> reweighting methods on systems that are more then a few residues. People
>> do use approximate reweighting schemes though, such as block averaging the
>> boost potentials over a small range of their simulation or taking a Taylor
>> expansion of the exponential and using that as a reweighting term. Of
>> course these are approximations to the true reweighting scheme, so they
>> should be used with caution.
>>
>> Jeff Wereszczynski
>> Assistant Professor of Physics
>> Illinois Institute of Technology
>> http://www.iit.edu/~jwereszc
>>
>>
>> On Wed, Nov 6, 2013 at 3:21 PM, Thomas Evangelidis <tevang3_at_gmail.com>wrote:
>>
>>> @James
>>>
>>> 1) Be cautious about the boost you apply, lipid bilayers are labile
>>> systems. Have a look at this paper for possible caveats and the effect of
>>> different threshold and alpha values:
>>>
>>> http://www.ncbi.nlm.nih.gov/pmc/articles/PMC3191728/
>>>
>>> As the authors state at the end, selective aMD is ideal for
>>> protein-membrane systems, however, it is not supported by NAMD. It will be
>>> though available in the next AMBER release.
>>>
>>>
>>> @Jeff
>>>
>>> A bit off-topic question: is it possible to calculate frequencies of
>>> properties like H-bonds from aMD trajectories by means of a reweighting
>>> procedure?
>>>
>>>
>>> On 6 November 2013 23:03, Jeff Wereszczynski <jwereszc_at_iit.edu> wrote:
>>>
>>>> Hi James,
>>>>
>>>> 1. I would agree that using dual boost is typically the best bet. As
>>>> you say, the general idea is that the torsional boost will increasing the
>>>> sampling of your biomolecule, while the total boost term will likely help
>>>> with diffusive properties.
>>>>
>>>> 2. Your values sound reasonable to me, if you have an average total
>>>> energy of -360,000 kcal/mol and 100k atoms then values of E=-340,000k and
>>>> alpha=20k might be a good place to start. If you want to increase sampling
>>>> I would try increasing E (for example, E=-320,000) or decreasing alpha.
>>>>
>>>> There have been a couple papers that have come out in the last year or
>>>> so looking at GPCR systems with aMD, I'm sure those will have more details
>>>> and you may find them interesting.
>>>>
>>>> Cheers,
>>>>
>>>> Jeff Wereszczynski
>>>> Assistant Professor of Physics
>>>> Illinois Institute of Technology
>>>> http://www.iit.edu/~jwereszc
>>>>
>>>>
>>>> On Wed, Nov 6, 2013 at 12:50 PM, James Starlight <
>>>> jmsstarlight_at_gmail.com> wrote:
>>>>
>>>>> Professor Wereszczynski,
>>>>>
>>>>> thank you for the explanations. Some of my suggestions:
>>>>> 1) Currently I'm interesting in the application of double boost to my
>>>>> protein-membrane system. As I understood the addition of the dihedrall
>>>>> bost can increase sampling mainly due to the increasing of the rotation
>>>>> (Sampling between +-60, 0 and 120 degrees states) in the phi and psi angles
>>>>> of polypeptide backbone as well as lipid tales (structural term). In
>>>>> addition the second (total boost, which depends on the <Utot> and the
>>>>> number of atoms) can increase sampling rate due to modifying diffusion
>>>>> rates of the solute into solvent. I suppose that for membrane-containing
>>>>> systems this could be especially significant because viscous lipids can
>>>>> significantly slave conformational dynamics of the protein increasing
>>>>> barriers between its different states.
>>>>> 2) Assuming <Utotal> ~ -360.000 (the most negative term is from the
>>>>> electrostatics) kcal/mol and the number of atoms is 100000 in my system I
>>>>> have obtained -340.000 Boost threshold as well as impirical value for
>>>>> alpha 20000 (100000/5). Than if I;d like to increase sampling I should a)
>>>>> decrease threshold and/or b) decrease alpha
>>>>> Does this statements correct?
>>>>>
>>>>> James
>>>>>
>>>>>
>>>>> 2013/11/6 Jeff Wereszczynski <jwereszc_at_iit.edu>
>>>>>
>>>>>> Hi James,
>>>>>>
>>>>>> If you wanted to calculate the energy of just the dihedrals in just
>>>>>> the protein, you could specify an atom selection in the NAMDEnergy plugin
>>>>>> to VMD to analyze just that part of your system.
>>>>>>
>>>>>> However, your boost energy is going to be applied to the entire
>>>>>> system and not just the protein, so the value you want to use is the total
>>>>>> dihedral energy for the system, which you already have calculated. Also,
>>>>>> the formula you quote is if you are applying a boost to all the energy
>>>>>> terms in the system. If you are just applying a boost to the dihedral
>>>>>> terms, you should use something more like this:
>>>>>>
>>>>>> E=<U_dihed>+4*number_residues
>>>>>> alpha=4/5*number_residues
>>>>>>
>>>>>> Thats just an initial guess, other people use different values. In
>>>>>> your case, the number_residues is the number of residues in your protein
>>>>>> and in your lipids. You can try those values and then if you are not
>>>>>> getting enough acceleration try adding the value of alpha you got to E.
>>>>>>
>>>>>> Cheers,
>>>>>>
>>>>>> Jeff Wereszczynski
>>>>>> Assistant Professor of Physics
>>>>>> Illinois Institute of Technology
>>>>>> http://www.iit.edu/~jwereszc
>>>>>>
>>>>>>
>>>>>> On Wed, Nov 6, 2013 at 1:29 AM, James Starlight <
>>>>>> jmsstarlight_at_gmail.com> wrote:
>>>>>>
>>>>>>> I try to specify my question dealing with more specified case:
>>>>>>>
>>>>>>> assuming I'm modelling activation of the membrane receptor (R->R*)
>>>>>>> which involves motion of the part of its helix which are not in the
>>>>>>> membrane but exposed to the solvent. So the energy for such conformation
>>>>>>> change (assuming that amplitude of such displacement could be ~ 3-5 A)
>>>>>>> could be ~ several kT. Therefore in equilibrium aMD I've never seen such
>>>>>>> transition due to big energy barrier between R and R*. Addition of the
>>>>>>> boost potential to dihedral term might solve this kinetic problem.
>>>>>>> so
>>>>>>>
>>>>>>> 1) firstly I need to compute the averaged Dihedral term for my
>>>>>>> protein only along the trajectory
>>>>>>> for the entire system with the VMD namdstats.tcl script (consisted
>>>>>>> of lipids as well) I have DIHED: 10068.803885714286
>>>>>>> How I could compute such value for only protein (excluding lipids)?
>>>>>>>
>>>>>>>
>>>>>>> 2) than assuming that E threshold = barrier height (several Kt) for
>>>>>>> the receptor of 350 atoms I have
>>>>>>>
>>>>>>> Eth= <Uav>+167 (according to the empirical formula U+1/3*number of
>>>>>>> atoms)
>>>>>>> alpha= 280 (4/5* number of atoms) > but not sure in this value
>>>>>>> because I'd like to preserve shape of the initial potential surface without
>>>>>>> its modification (only decreasing barrier height for the most stable (R)
>>>>>>> state= deeper potential well).
>>>>>>> So addition of such boost should increase transition rates across
>>>>>>> barrier in several times (I could not still understood the full meanings of
>>>>>>> the coefficient in that formulas). Does this statements correct?
>>>>>>>
>>>>>>> I'll be thankful for any suggestions.
>>>>>>>
>>>>>>> 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: thomasevangelidishomepage<https://sites.google.com/site/thomasevangelidishomepage/>
>>>
>>>
>>>
>>
>>
>> --
>> Jeff Wereszczynski
>> Assistant Professor of Physics
>> Illinois Institute of Technology
>> http://www.iit.edu/~jwereszc
>>
>>
>

This archive was generated by hypermail 2.1.6 : Tue Dec 31 2013 - 23:23:57 CST