AMBER Archive (2009)

Subject: [AMBER] RE: question

From: Ross Walker (ross_at_rosswalker.co.uk)
Date: Mon Nov 02 2009 - 17:36:48 CST

Hi Marek,

My apologies for not replying sooner. I was on travel at the beginning of
October and then came down with the flu which laid me out for 2 weeks and I
am only just beginning to get caught up.

> one of my colleague needs urgently to simulate interaction of the given
> protein with the solid surfaces (hydrophobic, hydrophilic) something
> similar like in
> attached article where was used Gromacs for this type of simulations.

In principal if it has been done in Gromacs it should be possible to do the
same thing in AMBER. It may need some tweaking to convert across the force
field parameters etc.
 
> Pavel already put the relevant questiones into Amber forum a few
> days/weeks
> ago but he got no answer until yet.
>
> Please see:
>
> http://archive.ambermd.org/200910/0017.html
> http://archive.ambermd.org/200910/0271.html
> http://archive.ambermd.org/200910/0291.html
>
> Moreover regarding to this topic, there are almost no information
> on the Amber forum (mostly just unaswered questions).
>
> So that's why I am disturbing you personally.
>
> The main question is of course if the current Amber version (10)
> is capable of this type of simulations or if we really have to move
> into
> Gromacs etc.

In principal AMBER 10 should be able to simulate such a system. The only
real limitation is in the fact that you cannot have bonds cross the periodic
boundaries so you will not be able to build an infinite covalently bonded
surface. You will have to cap the edges. I have done similar things myself
with cellulose surfaces without problems, just having weak restraints on the
corners to keep things in place.

It can be tricky to build such a system since Leap etc was designed for
proteins but it certainly is not impossible. It is just advanced and will
likely need some scripting of things to add in missing bonds etc. This is
certainly true if you have things cross bonded since leap only automatically
adds head and tail atoms so if you have more than 2 links per residue you
will have to use the bond command to add the missing ones. This will
obviously drive someone who doesn't know how to script crazy ;-).

The way to approach this will be to build a prepin file representing the
'unit' cell of your surface. Define head and tail atoms for each unit which
will automatically get bonded when you load the pdb. Then go through and add
the cross bonds etc. Realistically it is no different to building a protein
or a dendrimer for example.

> Maybe it is worth of mentioning that we can accept the approximation,
> that the surface atoms are fixed so the precise forcefield for the
> given
> surface (for example diamond) is not necessary, which could simplify
> the
> situation,
> however in principle would not be problem to find it somewhere and
> implemented in Amber using FRCMOD files.

If you really keep them fixed then this makes it a LOT simpler since you do
not need to worry about any of the bonds, angles or dihedrals for the
surface, just the EEL and VDW terms. Hence you should just build a residue
representing a unit of your surface. Do this in leap, add the charges and
atom types to get the correct VDW terms. Then do not bother to set head or
tail atoms. Save this as a lib file. Then edit your pdb and put each 'part'
of your surface as a separate residue, editing the residue number etc. Then
stick TER cards between it all and you should be able to load it up with
your protein on top fine.

> Although I already have ( unlike my colleague Pavel ) some experiences
> with
> Amber simulations, I have no clear idea how to parametrize given
> modified
> surface (using Antechamber) for the Amber calculation. More precisely,

This is a MUCH different question to. Can AMBER 10 simulate such a surface.
Getting the charges correct could be a Ph.D. project in itself. This is part
of the reason the paper you sent me is a PNAS one. I would look carefully at
how they parameterized it and follow that procedure. Exactly the same issue
is going to exist in Gromacs since you will have to give it charges and
parameters somehow so switching to Gromacs won't help you. I would suggest
doing a literature search on simulations of surfaces and see what people do
for charge fitting. This is well outside of what Antechamber was designed to
do so you are likely going to have to fit the charges yourself manually
using RESP. You will have to work out a way to cap these. If you have not
done much parameterization of surfaces before I would suggest looking for
people who have experience in this and see what approach they take.

> how
> to define the basic residui
> which could be ( after parametrisation in antechamber - creation of
> PREPIN
> and FRCMOD files )
> used for loading of the whole surface PDB file. I can just speculate
> that

I would find a way to generate the charges - likely manually using RESP.
Then create a pdb with your 'surface unit' in it, load it into leap and
build the residue manually, adding atom types, charges etc. I would not try
to use Antechamber or the GAFF force field for this since it is way outside
the design specs.

> If there is no standard "residui way" which could allow us to load PDB
> of
> the whole surface into tleap, I have just last idea to create simply
> just
> discrete array of the separate thiny pieces (like I described above).

Yes this is essentially what I am suggesting, find the smallest 'repeat'
unit for your surface and make this a residue. If you are keeping the
surface fixed it doesn't matter if it is not bonded properly etc. The key is
you will have to run QM charge fitting calculations on this so you will need
to make sure the valencies are complete etc.

> Of
> course this has some sense only in the case that such tiny piece of
> solid
> surface will be possible to parametrise in Antechamber as it is in case
> of
> "common" molecular structures.

Again, do not use Antechamber. Do this the 'old fashioned' and arguably MUCH
MORE RELIABLE way. Antechamber is a black box designed for screening
thousands of drug like molecules automatically. If you, or anyone else, are
working on a single system then you should ALWAYS do it by hand.

Good luck,
Ross

/\
\/
|\oss Walker

| Assistant Research Professor |
| San Diego Supercomputer Center |
| Tel: +1 858 822 0854 | EMail:- ross_at_rosswalker.co.uk |
| http://www.rosswalker.co.uk | PGP Key available on request |

Note: Electronic Mail is not secure, has no guarantee of delivery, may not
be read every day, and should not be used for urgent or sensitive issues.

_______________________________________________
AMBER mailing list
AMBER_at_ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber