Keep in mind that atom types are used for specific parts of the force
calculation. Consider each part and decide whether using either set of atom
types is best. For example, mixing atom types will make finding dihedral
combinations very difficult. The mixed versions won't be present, and you
need a way to derive them. Bonds and angles tend to be more transferable
than dihedrals, since theatter are fit after charge assignment, while bonds
and angles are fit before. My view is that you should mix as little as
possible, and only have 1 interface for the mixed model where you will have
to derive new parameters for the atom type combinations.
On Nov 2, 2016 9:55 AM, "Aditya G Rao" <aditya.grao.mail.huji.ac.il> wrote:
> Hi David,
>
> Thank you very much for the elaborate explanation! I followed your method
> and created a model of the cysteine dipeptide. I assigned the ff14SB atom
> types to the protein part (backbone and cysteine sidechain up to the
> sulphur) and gaff2 atom types to the modified side chain part (tetrapyrrole
> with ca 75 atoms). The charge on the unit is integral (-1).
>
> The derivation of the parameters is the starting point for my thesis, so I
> would like to get the best possible parameters.
>
> But I had a query regarding the atom type assignment. Do I assign the
> ff14SB atom types only to the cysteine part of the new non-standard residue
> or should do it also for the tetrapyrrole as much as possible? For example,
> there are methyl groups and propionate chains that already exist in ff14sb
> so I could assign them as well.
>
> After the assignment of the atom types, how do I go about solving the
> problem of charge fitting and bonded parameter derivation, such that I get
> parameters compatible with the ff14SB?
>
> Thanks in advance!
> Aditya
>
> On Thu, Oct 27, 2016 at 9:45 PM, David Cerutti <dscerutti.gmail.com>
> wrote:
>
> > Hi Aditya,
> >
> > I have some thoughts here. First, the tutorial on building modified
> > residues is pertinent to your case:
> >
> > http://ambermd.org/tutorials/basic/tutorial5/
> >
> > That will show you the critical step of building not just a ligand from
> > scratch but a residue which forms a link in the ongoing peptide chain.
> As
> > Carlos explained, antechamber / GAFF works when there are strictly
> > nonbonded interactions between the ligand and the rest of the protein,
> but
> > when there are bonds the valence parameters start to mix and the trouble
> > comes with bookkeeping and conventions (this is not so much a limitation
> of
> > antechamber / GAFF as a limitation of any system that would attempt to do
> > something like that). The tutorial, however, shows you how to use
> prepgen
> > (see Section 3 on that page) to spell out what happens when this modified
> > Cysteine becomes part of a chain, or even if it were to appear at the N-
> or
> > C-terminus. Once you get over that step, it becomes smoother sailing.
> >
> > You'll have a prepi file coming out of that tutorial, which is something
> > you can then read into tleap equivalent to any residue library file.
> > Antechamber / GAFF will have generated the charges and atom types, and
> the
> > prepgen step will have made it so that this thing can be trimmed of what
> > would be leaving groups (N-terminal hydrogen and C-terminal hydroxyl) in
> > the actual chemistry of putting this thing into a growing peptide chain.
> > However, the charges are going to stay as they were from the whole
> molecule
> > case you fed to antechamber, and you've wisely asked for parameters that
> > are compatible with ff14SB.
> >
> > Here's how I would do that:
> > - Proceed to make the main chain residue .prepi file using the steps in
> > Tutorial B5 above.
> > - Modify the prepi file to alter the atom types on the amino acid to
> ff14SB
> > types.
> > - You're going to change a lot of lowercase letters to uppercase.
> > - Use ${AMBERHOME}/dat/leap/lib/amino12.lib, CYS as a guide.
> > - Create, in tleap, the system Ace-(Modified Cys)-Nme. Here's my guess
> as
> > to the input:
> >
> > source leaprc.protein.ff14SB
> > loadAmberPrep (Your .prepi file for modified Cys)
> > loadAmberParams (.frcmod file you will create and edit until this works)
> > ModCysDi = sequence { ACE (Your Modified Cys Residue Code) NME }
> > saveAmberParm ModCysDi ModCysDipeptide.top ModCysDipeptide.crd
> > quit
> >
> > - Above, please don't be scared that I said "frcmod file you will create
> > and edit until this works." That's what's going to mitigate and define
> the
> > scope of parameters you have yet to solve, to link GAFF atom types to
> > ff14SB atom types. When you first run the input above, because you have
> > modified the amino acid's main chain atom types, you've introduced a
> break
> > point at which it doesn't know "dihedral between GAFF type 1, GAFF type
> 2,
> > ff14SB SH, ff14SB 2C" and so forth. This sort of parallels the stuff the
> > tutorial has you doing with making two other frcmod files. Just manually
> > construct your frcmod, every time it complains about not knowing a
> > parameter, add that parameter and take the values from something similar
> > you find in gaff.dat or parm10.dat, or put placeholder values of zero.
> > - Now, you've got a topology that includes your modified residue AS A
> > DIPEPTIDE. This is how everything in the Cornell / Simmerling force
> fields
> > (including ff14SB) gets parameterized, pretty much.
> > - For things that are truly compatible with ff14SB, you will want to keep
> > the backbone N, H, C, and O charges consistent with the rest of the main
> > chain non-ionic amino acids, and "for best results" you should make
> > appropriate modifications after that atom clipping business with
> > prepgen--I'm not sure it would even leave you with an integer charge on
> the
> > residue, let alone neutrality. There may be a way to solve the new
> charges
> > with the RED server, but the way I would do it is with mdgx. Once you
> get
> > to this stage, the point where you have a working dipeptide of the
> molecule
> > with an actual topology, the routines I've designed for charge fitting
> and
> > bonded parameter derivation can do all sorts of things, including mimic
> the
> > protocol for the ff95 ... ff99 ... ff14SB lineage.
> >
> > Let me know when you get to this stage and are able to make Ace-(modified
> > Cys)-Nme with ff14SB atom types on the protein part and GAFF atom types
> > remaining on the modified side chain part. This shows that you've
> > conquered the bookkeeping part of the problem and are ready to smooth out
> > the representation with "good" numbers. Depending on whether this is a
> > one-off quick simulation or the basis for thesis work, I can advise you
> on
> > manual parameter selection or crunching the numbers with some cluster
> time
> > to explicitly derive each charge and force constant.
> >
> > Dave
> >
> >
> > On Thu, Oct 27, 2016 at 1:31 PM, Aditya G Rao <
> aditya.grao.mail.huji.ac.il
> > >
> > wrote:
> >
> > > Dear Carlos,
> > > Thank you. So is it better to use Amber ff14SB for the entire
> > non-standard
> > > residue to ensure consistency with the protein that it has been
> connected
> > > with?
> > >
> > > Also could I use RED program to generate parameters consistent with the
> > > ff14SB force field?
> > >
> > > Thanks in advance!
> > >
> > > On Thu, Oct 27, 2016 at 7:38 PM, Carlos Simmerling <
> > > carlos.simmerling.gmail.com> wrote:
> > >
> > > > The problem you will encounter is at the boundaries between gaff and
> > > > non-gaff atom types. For example, you need a bond parameter linking
> > atoms
> > > > from 2 different models (gaff and ff14sb) , and it likely won't
> exist.
> > > The
> > > > same goes for using gaff on a whole residue that must be bonded to
> > > residues
> > > > that use other force fields. Currently gaff works well when
> > interactions
> > > > between the 2 force fields are only between nonbonded interactions.
> If
> > > you
> > > > want to do what you propose, you will probably need to create a
> frcmod
> > > file
> > > > with all of the mixed terms for bonds, angles and dihedrals at the
> > > > interface. This can be done, but depending on what you are modeling
> it
> > > may
> > > > or may not already exist.
> > > >
> > > > On Oct 27, 2016 12:21 PM, "Aditya G Rao" <
> aditya.grao.mail.huji.ac.il>
> > > > wrote:
> > > >
> > > > Dear Amber users,
> > > > I'm trying to parameterise a non-standard residue. The residue is
> > > composed
> > > > of a cysteine and a tetrapyrrole bound to the sulphur of this
> cysteine.
> > > >
> > > > My question is whether I can combine gaff2 and Amber (e.g. ff14sb)
> > > > parameters for this non-standard residue? I could use ff14SB for the
> > > > cysteine that is well parametrized and gaff2 for the tetrapyrrole. Is
> > it
> > > > necessary to use only gaff2 or only Amber parameters for the whole
> > > residue?
> > > >
> > > > Thanks in advance!
> > > > Aditya
> > > >
> > > > --
> > > > Aditya G. Rao
> > > > PhD Student
> > > > Fritz Haber Research Center for Molecular Dynamics
> > > > Institute of Chemistry, HUJI
> > > > Givat Ram, Jerusalem, Israel
> > > > _______________________________________________
> > > > AMBER mailing list
> > > > AMBER.ambermd.org
> > > > http://lists.ambermd.org/mailman/listinfo/amber
> > > > _______________________________________________
> > > > AMBER mailing list
> > > > AMBER.ambermd.org
> > > > http://lists.ambermd.org/mailman/listinfo/amber
> > > >
> > >
> > >
> > >
> > > --
> > > Aditya G. Rao
> > > PhD Student
> > > Fritz Haber Research Center for Molecular Dynamics
> > > Institute of Chemistry, HUJI
> > > Givat Ram, Jerusalem, Israel
> > > _______________________________________________
> > > AMBER mailing list
> > > AMBER.ambermd.org
> > > http://lists.ambermd.org/mailman/listinfo/amber
> > >
> > _______________________________________________
> > AMBER mailing list
> > AMBER.ambermd.org
> > http://lists.ambermd.org/mailman/listinfo/amber
> >
>
>
>
> --
> Aditya G. Rao
> PhD Student
> Fritz Haber Research Center for Molecular Dynamics
> Institute of Chemistry, HUJI
> Givat Ram, Jerusalem, Israel
> _______________________________________________
> AMBER mailing list
> AMBER.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber
>
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Wed Nov 02 2016 - 07:30:02 PDT
