Re: [AMBER] [q4md-fft] General Question about Charge Derivation using RESP

From: FyD <fyd.q4md-forcefieldtools.org>
Date: Wed, 24 Aug 2011 08:41:58 +0200

Dear Farid Ismail,

> One question I have, is how do I tell whether, RESP, ESP, or
> Connolly is the 'best'? Put it another way, is there a standard
> that people use to tell this? Compare to physical properties (but
> what if the physical properties doesn't converge)?

I would answer try & see ;-)

In R.E.DD.B., you can already find such a comparison:
http://q4md-forcefieldtools.org/REDDB/download.php

Search project(s) by
-- R.E.DD.B. code (if known)
-- Molecule keyword x
-- Molecule name
-- Author lastname
-- Theory level/Basis set
          Text Organic

Search... [Done]
Result(s) for search by Molecule keyword Organic

- http://q4md-forcefieldtools.org/REDDB/projects/W-46/
RESP atomic charges for 10 solvent molecules; CONNOLY SURFACE

- http://q4md-forcefieldtools.org/REDDB/projects/W-47/
RESP atomic charges for 10 solvent molecules; CHELPG

- http://q4md-forcefieldtools.org/REDDB/projects/W-48/
ESP atomic charges for 10 solvent molecules; CONNOLY SURFACE

- http://q4md-forcefieldtools.org/REDDB/projects/W-49/
ESP atomic charges for 10 solvent molecules; CHELPG

More generally, we tried to summarize most of parameters that affect
non-polarizable charge values in the R.E.D. publication (PCCP 2010).
This can help you to test even more possibilities... For instance, you
could also try the Geodesic algorithm by M. Spackman that is available
in GAMESS/Firefly (just modify the GAMESS input in the R.E.D. source
code)...

regards, Francois


> Dear Ismail,
>
>> I want to perform organic solvent simulation using Amber. I built
>> my test system dimethoxy ethane (CH3-O-CH2-CH2-O-CH3). I have tried
>> to get the charges using RED-vIII.4. I tried a few conformations,
>> then check the density with Amber and compare to the experimental
>> density.
>
> - For dimethoxy ethane, let's start with chemical equivalencing:
> see http://q4md-forcefieldtools.org/REDS/news.php
> http://q4md-forcefieldtools.org/REDS/popup/popanteredtopequiv.php
>
> Chemically equivalent atoms should bear the same charge values in MD
> simulations. Ante_R.E.D. 1.x does not rigorously deal with chemical
> equivalencing, while Ante_R.E.D. 2.0 in R.E.D. Server does.
>
> In your case, Ante_R.E.D. 1.x will generate the following atom names
> (in the first column of atom names in the P2N file format):
> CT1 H1 H1 H1 O2 CT3 H3 H3 CT4 H4 H4 O5 CT6 H6 H6 H6
> this has to be modify by hand
> (see the tutorials . http://q4md-forcefieldtools.org/Tutorial/Tutorial-1.php)
>
> Ante_R.E.D. 2.0 will generate:
> CT1 H1 H1 H1 O2 CT3 H3 H3 CT3 H3 H3 O2 CT1 H1 H1 H1
> that is correct
>
> About the P2N file format, atom naming and chemical equivalencing see:
> http://q4md-forcefieldtools.org/Tutorial/Tutorial-1.php#3
> http://q4md-forcefieldtools.org/Tutorial/Tutorial-1.php#ATOM-NAME
> http://q4md-forcefieldtools.org/Tutorial/Tutorial-1.php#RULES
>
>
> - You can find ethers in R.E.DD.B. as examples:
> http://q4md-forcefieldtools.org/REDDB/download.php
>
> Search project(s) by
> -- R.E.DD.B. code (if known)
> -- Molecule keyword x
> -- Molecule name
> -- Author lastname
> -- Theory level/Basis set
> Text Ether
>
> Search... [Done]
> Result(s) for search by Molecule keyword ether
> Project name Methoxyethane
> Project code W-17
>
> etc...
>
> http://q4md-forcefieldtools.org/REDDB/projects/W-17/
> & corresponding RESP inputs:
> http://q4md-forcefieldtools.org/REDDB/projects/W-17/input1.in
> http://q4md-forcefieldtools.org/REDDB/projects/W-17/input2.in
>
> (Obviously chemical equivalecing in 'methoxyethane' is different to
> that in 'dimethoxy ethane')
>
>> My questions are:
>> (1) I include a few conformations with my calculation.
>
> 'Few' conformations is not accurate - a set of conformations is
> involved in charge derivation - if the difference of energy between
> the conformations, dE is 'small'; < 2-4 kcal/mol? Thus, in the Amber
> FF, one or two conformations are often involved in charge derivation.
> Please, see:
> http://onlinelibrary.wiley.com/doi/10.1002/jcc.540161106/abstract
>
>> Under certain set of conformations, the resulting charges are
>> different for similar atoms. The molecule dimethoxy ethane is
>> symmetic.
>
> See chemical equivalencing above.
>
>> Why does this happen? Is it OK to just average the charges out
>> between similar atom?
>
> 'charge fitting' is not 'charge averaging':
> - by charge fitting you involve multiple conformations (and multiple
> orientations) to generate a single set of charge values
> - by charge averaging you generate multiple sets of charge values
> based on multiple conformations (taken individually) and you average
> them to generate a single set of charge values.
>
>> (2) Sometimes, when use more conformation, I get worse density
>> value compare to experimental value. For example, I include just
>> the gauche conformation at O-CH2-CH2-O and get density of 0.9000
>> (experimental value is 0.8683). If I use the gauche and the trans
>> conformation, I get 0.8900. I thought if I include more
>> conformations like tTt, gTg, etc (the lower case is for the
>> CH3-O-CH2-CH2) I get density ~0.8950, which is worse. Is this
>> normal? I assume then it is safe if I use only the two conformation
>> values?
>
> 'more' is vague once again; I would first try to select a conformation
> that makes sense i. e. the lowest minimum or the two lowest minima.
> etc...
>
> - I would also try ESP charge derivation instead of RESP charge derivation.
> - I would try the Connolly surface or the CHELPG algorithm in MEP
> computation as well; please see:
> http://q4md-forcefieldtools.org/REDS/popup/popkeyword.php
>
>> (3) I don't quite get the REMARK REORIENT part. Does that part
>> mean RESP is holding certain atoms constant with respect to certain
>> axes?
>
> The orientation of an optimized geometry affect MEP based charge
> values. We observed a max. charge uncertainty of .07 e for a carbon
> atom. See
> http://www.ncbi.nlm.nih.gov/pmc/articles/PMC2918240/
> http://www.ncbi.nlm.nih.gov/pmc/articles/PMC2918240/#S7
>
> The main idea with the rigid-body reorientation algorithm (RBRA)
> implemented in the R.E.D. program is to derive reproducible charge
> values independently of the QM program used (GAMESS, Firefly or
> Gaussian) and independently of the initial orientation chosen by the
> user. This is achieved by rigorously controlling the orientation of
> the optimized geometry before MEP computation.
>
> In general, we use pairs of molecular orientations; we select three
> atoms in the RBRA algorithm, X Y Z to define a first orientation, and
> then Z Y X in a second one to cancel out the effect of the first
> orientation by the second one.
>
>
> I hope this helps. Do not hesitate to ask for more information if
> something is not clear. I forward this message in the Amber mailing
> list as people there might be interested in this topic.
>
> regards, Francois



_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Wed Aug 24 2011 - 00:00:02 PDT
Custom Search