From: Gordon Wells (gordon.wells_at_gmail.com)
Date: Wed May 09 2012 - 10:04:04 CDT
argh, iirc gmail's mailing list detection used to be better than this
-- max(∫(εὐδαιμονία)dt)
Dr Gordon Wells
Chemistry Department
Emory University
Atlanta, Georgia, USA
---------- Forwarded message ----------
From: Gordon Wells <gordon.wells_at_gmail.com>
Date: 9 May 2012 10:54
Subject: Re: AW: namd-l: stereo-chemical inversion during MD
To: Branko <bdrakuli_at_chem.bg.ac.rs>
Hi Branko
Thanks for the suggestions. When I first saw this I did include those water
molecules too (without them some waters do soak into the same places). But
I see the same thing when setting up a simulation of the ligand in water
only. I think in that particular conformation there isn't enough of a
screening effect (or competitive forces) between the carboxyl and ammonium
groups.
-- max(∫(εὐδαιμονία)dt)
Dr Gordon Wells
Chemistry Department
Emory University
Atlanta, Georgia, USA
On 8 May 2012 16:03, Branko <bdrakuli_at_chem.bg.ac.rs> wrote:
> Gordon,
>
> I have no idea and never seen such mobility, especially of the -NH3 after
> the inversion. Axel most probably can offer some useful advice. In 2A5T
> there is 3 water molecules that makes bridged H bonds: NH3 (HOH 2019) and
> terminal COOH (HOH 2017 and 2004). Maybe you can try to include this water
> molecules (I could suggests GRID and -6 kcal/mol as reasonable for exact
> positioning, see J. Chem. Inf. Model. 2011, 51, 2860–2867). Also try to
> check pronation state of your ligand by Propka (http://propka.ki.ku.dk/),
> vicinal residues can dramatically changes pKa, and propka uses empirical
> function for estimation. If you prepare protein-ligand complex for Desmond
> by Maestro, and use protein preparation wizard, one step include
> optimization of the H-bonds network, so this also can make difference.
>
> Branko
>
>
> On 5/8/2012 9:04 PM, Gordon Wells wrote:
>
> Hi
>
> Thanks for the helpful replies.
>
> It is indeed a case of overcoming a high energy barrier and it seems to
> be primed by the starting conformation, first minimisation squashing the
> bond angle (I accidentally included both conformations in the second
> attachment above), and then large initial forces during relaxation (after
> reading the paper Francesco cited). I've attached an .mpg of the first few
> frames during equilibration in the protein-ligand complex (I tried starting
> the relaxation at 10K and 60K). While based on a homology model, the l-glu
> (ligand) conformation is exactly as in the template (2A5T). This protein
> family also binds d-glu and I've observed inversion for that ligand too.
>
> It seems adding restraints during the relaxation stage with the
> chirality plugin is needed here. But I'm still curious as to what the other
> MD packages I tried are doing differently?
>
> Regards
> Gordon
>
>
> -- max(∫(εὐδαιμονία)dt)
>
> Dr Gordon Wells
> Chemistry Department
> Emory University
> Atlanta, Georgia, USA
>
>
>
> On 8 May 2012 11:51, Branko <bdrakuli_at_chem.bg.ac.rs> wrote:
>
>> Axel,
>>
>> This is valuable information, still I didn't faced with similar situation
>> so far. Also, I can suppose than something like this is possible for sp3 N,
>> still cannot imagine such high energy intermediate of sp3 C. On the other
>> hand Gordon simulate one amino acid in explicit solvent and obtain
>> stereochemical inversion. Did he start with conformer of unrealistic high
>> energy. In protein, surroundings can constrain this residue in such
>> conformation, still think that something in initial settings did not work
>> perfectly. Maybe high temperature or similar, also there is no information
>> about origin of the starting protein, taken from the PDB, modeled. Gordon
>> state that can overcome problem by 'increasing the di-electric or
>> decreasing the partial charges on the ammonium hydrogens'. Is there some
>> metal ions in vicinity. Also are free amino acid is comparable with the
>> same one bound in the protein; even if this is C terminal amino acid 'NH2'
>> should be part of amide bond - so this isn't NH3+...
>>
>> Branko
>>
>>
>>
>>
>>
>> On 5/8/2012 3:33 PM, Axel Kohlmeyer wrote:
>>
>> On Tue, May 8, 2012 at 4:26 AM, Branko <bdrakuli_at_chem.bg.ac.rs> <bdrakuli_at_chem.bg.ac.rs> wrote:
>>
>> Stereochemical inversion implies bond breaking and bond making, and this is
>> not possible by molecular dynamics and using all atoms force fields. It is
>>
>> i beg to differ. stereochemical inversion is quite possible
>> without bond breaking and particularly with a classical
>> model. all you need is to go through a very high-energy
>> intermediate. and particularly this last condition is what
>> makes it particularly possible for a classical model, since
>> bonds do *not* break in this case. when setting up a
>> simulation and trying to get it to equilibrium, these kind
>> of conditions can happen (and they happen more frequently
>> than people suspect).
>>
>> just consider the case of a asymmetrically substituted
>> methane. all you need is a conformation where all substituents
>> are in the same plane as the central carbon atom. depending
>> on how the planar symmetry is broken, you will get a
>> different stereo isomer.
>>
>> VMD ships with two plugins that help to detect such artefacts:
>> http://www.ks.uiuc.edu/Research/vmd/plugins/chirality/http://www.ks.uiuc.edu/Research/vmd/plugins/cispeptide/
>>
>> cheers,
>> axel.
>>
>> advisable to carefully check your input PDB file, residue name, atoms etc.
>>
>>
>> On 5/8/2012 8:01 AM, Norman Geist wrote:
>>
>> Hi Gordon,
>>
>>
>>
>> I’m not a specialist for biochemistry, but what you say could make sense as
>> the minimization looks for the best energy conformation. It could be, that
>> this angle is a better energy conformation for a particular system. But when
>> it’s bounded to other residues, it would maybe interfere with other side
>> chains and so would stay with the original angle. That would be easy to try
>> out. I can hardly imagine that the minimization algorithm can work against
>> the FF parameters.
>>
>>
>>
>> Norman Geist.
>>
>>
>>
>> Von: owner-namd-l_at_ks.uiuc.edu [mailto:owner-namd-l_at_ks.uiuc.edu <owner-namd-l_at_ks.uiuc.edu>] Im Auftrag
>> von Gordon Wells
>> Gesendet: Montag, 7. Mai 2012 19:01
>> An: namd-l_at_ks.uiuc.edu
>> Betreff: namd-l: stereo-chemical inversion during MD
>>
>>
>>
>> Hi All
>>
>>
>>
>> I've encountered a strange situation when simulating a particular
>> conformation of L-Glu. During minimisation the bond angle between the
>> carboxyl-C, C-alpha and amino-N decreases from 112° to 89°. When this is
>> subsequently used for MD there is often a stereo-chemical inversion around
>> the C-alpha. I see this when simulating the system in its original protein
>> complex and free in solution (TIP3 solvent for both).
>>
>>
>>
>> I can prevent it by using a very short minimisation (50 instead of 1000
>> steps), increasing the di-electric or decreasing the partial charges on the
>> ammonium hydrogens. Nonetheless, I'm sure this strained conformation
>> shouldn't be produced in the first place (I'm not able to replicate this
>> behaviour in macromodal or desmond) The force between the carboxyl oxygen
>> (nearest to the ammonium moiety) and the ammonium hydrogens seems to be too
>> high.
>>
>>
>>
>> I've attached before and after pdbs of the free L-glu. I get the distorted
>> conformation from the following namd input (with and without pbc):
>>
>>
>>
>> coordinates LGlu_autopsf.pdb
>>
>> structure LGlu_autopsf.psf
>>
>>
>>
>> paratypecharmm on
>>
>> parameters par_all27_prot_lipid_na.inp
>>
>>
>>
>> outputname minall-lglu-only
>>
>> binaryoutput yes
>>
>> outputenergies 25
>>
>>
>>
>> switching on
>>
>> cutoff 12
>>
>> switchdist 10
>>
>> pairlistdist 14
>>
>> exclude 1-4
>>
>>
>>
>> fixedAtoms off
>>
>>
>>
>> numsteps 1000
>>
>> dielectric 1
>>
>> minimization on
>>
>>
>>
>> Is this forcefield related, bad input file or possibly a bug in NAMD?
>>
>>
>> -- max(∫(εὐδαιμονία)dt)
>>
>> Dr Gordon Wells
>> Chemistry Department
>>
>> Emory University
>>
>> Atlanta, Georgia, USA
>>
>>
>>
>>
>>
>> No virus found in this incoming message.
>> Checked by AVG - www.avg.com
>> Version: 9.0.930 / Virus Database: 2410.1.1/4983 - Release Date: 05/07/12
>> 08:35:00
>>
>>
>>
>>
>>
>>
>>
>> No virus found in this incoming message.
>> Checked by AVG - www.avg.com
>>
>> Version: 9.0.930 / Virus Database: 2410.1.1/4984 - Release Date: 05/07/12 20:34:00
>>
>>
>>
>>
>
>
>
> No virus found in this incoming message.
> Checked by AVG - www.avg.com
>
> Version: 9.0.930 / Virus Database: 2410.1.1/4985 - Release Date: 05/08/12 08:41:00
>
>
>
>
This archive was generated by hypermail 2.1.6 : Mon Dec 31 2012 - 23:21:31 CST