On Thu, 2014-09-18 at 12:39 +0300, Thomas Evangelidis wrote:
> On 17 September 2014 18:42, Jason Swails <jason.swails.gmail.com> wrote:
>
> > On Wed, Sep 17, 2014 at 11:33 AM, Thomas Evangelidis <tevang3.gmail.com>
> > wrote:
> >
> > > On 16 September 2014 19:13, Jason Swails <jason.swails.gmail.com> wrote:
> > >
> > > > On Tue, 2014-09-16 at 17:05 +0300, Thomas Evangelidis wrote:
> > > > > Greetings,
> > > > >
> > > > > I want to run MMPBSA.py script using 3 different trajectories
> > (complex,
> > > > > receptor, ligand). The difference between the receptor and the
> > complex
> > > > > trajectories was that the receptor had an aspartate protonated which
> > > was
> > > > > unprotonated in the complex. This is perfectly fine in reality, but
> > > when
> > > > it
> > > > > attempted to use MMPBSA.py the program complained.
> > > >
> > > > This is actually not fine (at least not according to the details you've
> > > > written here). The MM/PBSA thermodynamic cycle does not accommodate
> > > > "chemical" changes (i.e., changes in covalent bonds). There is an
> > extra
> > > > process of deprotonation and proton desolvation that is not accounted
> > > > for. So you would need to add on a thermodynamic cycle similar to the
> > > > kind used for pKa calculations and constant pH simulations.
> > > >
> > > > > So I modified the
> > > > > receptor trajectory and topology by removing the extra proton from
> > the
> > > > ASP
> > > > > side-chain. However, I still get a residue mismatch error that may be
> > > > cause
> > > > > by the modification I introduced. Do you believe this is the case? If
> > > > yes,
> > > > > is there any work around to fix it?
> > > >
> > > > MMPBSA.py takes steps to catch errors that break the assumptions of
> > > > MM/PBSA analyses (like what you are doing here). [1] Simply stripping
> > > > the extra proton will not work -- MMPBSA.py also checks that the
> > residue
> > > > sequences are the same in the bound and unbound states (ASH != ASP) and
> > > > it checks that the atoms have the same charges in the bound and unbound
> > > > states... which they won't if all you did was strip out a hydrogen atom
> > > > using cpptraj.
> > > >
> > > > You can use MMPBSA.py to process each trajectory independently and then
> > > > combine them yourself as legs of the applicable thermodynamic cycle for
> > > > your process.
> > > >
> > > >
> > > >
> > > Good point. The protein is a transporter with a binding pocket that
> > becomes
> > > solvent exposed (outward facing conformation) when ASP51 is protonated
> > > (ASH51). So the receptor trajectory was actually created to monitor the
> > > opening of the substrate binding cavity. Deprotonation of ASH51 occurs
> > upon
> > > ligand binding and triggers the occluded ==> inward facing transition. My
> > > purpose is to compare free energies of binding for a series of analogues
> > > with know kinetics. This protonation-deptonation is associated with large
> > > conformational changes and hence a significant entropic change which is
> > > impossible to estimate. So I thing my best bet is to calculate relative
> > > free energies between the ligands (given that the entropic cost of
> > binding
> > > is quite similar for each of them) by using only the complex
> > trajectories I
> > > have.
> > >
> >
> > Sounds reasonable.
> >
> > I tried to do that for one complex but I am getting a series of errors of
> > > the following type:
> > >
> > > PB Bomb in pb_aaradi(): No radius assigned for atom 65 CB 3C
> > > PB Bomb in pb_aaradi(): No radius assigned for atom 79 CB 2C
> > > PB Bomb in pb_aaradi(): No radius assigned for atom 115 CG CO
> > >
> > > These are associated with protein atoms, which is strange since I used
> > > ff14sb. I tried to correct them with parmed.py using:
> > >
> > > change AMBER_ATOM_TYPE .%CB C
> > > change AMBER_ATOM_TYPE @%CG C
> > >
> >
> > It looks to me like you are substituting atom names for types. I think
> > this should read
> >
> > change AMBER_ATOM_TYPE .%3C CT
> > change AMBER_ATOM_TYPE .%2C CT
> > change AMBER_ATOM_TYPE @%CO CT
> >
> >
> > > Is this correct? Should all these carbons be of type 'C' ?
> > >
> > >
> >
> > I would use type CT (used to be the prototypical sp3-hybridized C used for
> > all Calpha's in ff99SB and earlier). C is the carbonyl type, I believe.
> > Of course it may be safer to check the parameter files and see what the
> > 3C, 2C, and CO types really are and pick the closest-matching type from
> > ff99SB. Although I think you can specify radiopt=0 to use the radii in the
> > prmtop file instead? (it doesn't appear that PBSA has updated its atom type
> > lists to support the new types in ff14SB).
> >
> >
> >
> Yes, radiopt=0 suppresses all these errors.
>
> I have two more queries. Since my protein is transmembrane, I defined the
> receptor to include both the protein and the membrane. However, the
> membrane spans the central box only. Does MMPBSA.py take into account the
> periodic membrane bilayers at the x-y plane as well? I get "UserWarning:
> Solvated topology 0 has IFBOX == 0" because I stripped the water when I
> concatenated all individual trajectories. Does this prevent MMPBSA.py from
> accounting PBC?
MMPBSA.py never accounts for PBC. There is only one GB implementation
that I'm aware of that does (and I don't know how much sense such an
approach makes).
You may want to look into the membrane capabilities of PBSA. It's
relatively new functionality, but it was designed for studies like this
(it treats the membrane and solvent implicitly by setting up multiple
dielectric regions).
> And the second query has to do with mpi. I don't have the AMBER14 source
> yet, so I cannot compile AMBER-MPI version (AMBER12 fails to compile with
> AmberTools14). So, although serial execution of MMPBSA.py runs smoothly,
> when I invoke it with mpirun it runs multiple times the same calculation.
> Does this have to do with compilation of AMBER14 source?
You need to invoke MMPBSA.py.MPI with mpirun. MMPBSA.py is strictly
serial. The former is built with AmberTools 14 (as long as you build
AmberTools 14 in parallel).
Hope this helps,
Jason
--
Jason M. Swails
BioMaPS,
Rutgers University
Postdoctoral Researcher
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Thu Sep 18 2014 - 05:00:09 PDT
