Re: Generalized Born Solvent Question

From: Aron Broom (broomsday_at_gmail.com)
Date: Fri Aug 03 2012 - 15:21:22 CDT

Hi Roy,

Yes, when doing GBIS, there is an option "SASA" which is "off" by default,
meaning that after doing the GBIS electrostatics, the VDW and other forces
are then calculated and dynamics are performed. If you stipulate "SASA
on", an additional force is applied based on solvent accessible surface
area calculations for each atom, which has the effect of attempting to bury
solvent accessible atoms. In theory, this should make GBIS calculations
more accurately match explicit solvent calculations.

It's been pointed out to me by various implicit solvent experts, that
without the SASA correction, GBIS still tends to recapitulate behaviours
such as proteins burying hydrophobic groups in their cores, because the
breaking of weak hydrophobic VDW interactions are not compensated by any
solvent VDW interactions. Nevertheless, from what I've seen with both
ligand binding and protein folding simulations, applying the SASA
correction is fairly critical for getting the correct energies from free
energy calculation methods, and also gets protein structures that more
accurately reflect known crystal or solution structures (my own experience
anyway).

In AMBER 11 the SASA correction wasn't supported when using the GPU, but it
is now in AMBER 12. In NAMD 2.9 if one does normal GBIS on say a 2000 atom
protein (~150 amino acids) the speed on a typical M2070 (GTX570 equivalent)
GPU is ~15 ns/day, but adding in the SASA calculation drops it to ~5
ns/day, which I believe is because the SASA calculations are still
performed on the CPU, whereas the GBIS calculations recently (after NAMD
2.8) migrated to the GPU. I haven't had a chance to play with AMBER 12, so
I can't actually be certain that with everything on the GPU the SASA
corrected simulation will be nearly the same speed as without, but my
suspicion is that it will.

As far as documentation, it's in the 2.9 user guide. Just open the PDF and
do a search for GBIS, it should bring you to the correct section. They
reference in their the LCPO (Linear Combination of Partial Overlaps) method
that is used for calculating SASA. This is the same method that AMBER
uses, but I believe GROMACS uses a different method. I'm not sure which
method VMD uses when you invoke a SASA calculation through the TK console,
probably that is in the relevant documentation also.

If you're looking for the SASA source code, I was peeking into it recently,
and I think the c file is named in such a way that it's hard to miss or
confuse with other codes.

~Aron

On Fri, Aug 3, 2012 at 3:07 PM, Roy Fernando <roy.nandos_at_gmail.com> wrote:

> Hi Aron,
>
> Did you refer to "Solvent Accessible Surface Area" by SASA?
>
> Can we do this measurement in NAMD, in GBIS? I am intrigued!
>
> Is there any documentation about using SASA in GBIS?
>
> Regards,
>
> Roy
>
>
>
>
>
>
> On Mon, Jul 30, 2012 at 6:06 PM, Aron Broom <broomsday_at_gmail.com> wrote:
>
>> Just as an FYI to anyone who has this same question and discovers this
>> thread, Dr. David Case informs me that once the Generalized Born
>> calculations have taken place, all of the atoms are effectively chargeless,
>> since the electrostatic portion of the calculations are done, and
>> therefore, it is safe to treat SASA equally for all of them.
>>
>> ~Aron
>>
>>
>> On Thu, Jul 19, 2012 at 2:34 PM, Aron Broom <broomsday_at_gmail.com> wrote:
>>
>>> Hi Everyone,
>>>
>>> I've asked before about the SASA term with GBIS simulations, but I'm
>>> still slightly confused and have what I think is a rather quick question:
>>>
>>> So I understand that in a GBIS simulation without SASA, the lack of
>>> potential VDW interactions between water and say a hydrophobic sidechain,
>>> means that the sidechain will tend to favour interacting with the protein,
>>> thereby giving a partial hydrophobic effect. Moreover, when one adds in
>>> the SASA term, the energy gradients tend to favour burial of exposed
>>> surfaces, ideally, recapitulating in full the hydrophobic effect. What I
>>> don't understand, is that in my reading of the articles presenting these
>>> methods, I don't see anything about the polarity of the atoms being taken
>>> into account for this SASA term. To me, this would suggest that a solvent
>>> accessible polar group is just as unfavourable as a similarly exposed
>>> hydrophobic group as far as SASA is concerned. Furthermore, because
>>> enthalpic interactions between polar groups are stronger than between
>>> hydrophobic groups, this would imply that if one ran a GBIS + SASA
>>> simulation of a protein for long enough, you'd end up with a "native state"
>>> in which the polar groups were all buried in the core.
>>>
>>> There must be something critical I'm missing here?
>>>
>>> ~Aron
>>>
>>> --
>>> Aron Broom M.Sc
>>> PhD Student
>>> Department of Chemistry
>>> University of Waterloo
>>>
>>>
>>
>>
>> --
>> Aron Broom M.Sc
>> PhD Student
>> Department of Chemistry
>> University of Waterloo
>>
>>
>

-- 
Aron Broom M.Sc
PhD Student
Department of Chemistry
University of Waterloo

This archive was generated by hypermail 2.1.6 : Mon Dec 31 2012 - 23:21:53 CST