On Fri, Jun 28, 2013 at 4:22 AM, Hannes Loeffler <Hannes.Loeffler.stfc.ac.uk
> wrote:
> I have not tested against AmberTools 12 yet but I've also got a dG which
> seems too high. I have however checked with mm_pbsa.pl and the results
> are totally different: DELTA TOTAL = -9.6205 vs PBTOT = -47.07.
> The protocols are not fully compatible I suppose but I add the relevant
> output below.
>
> [snip]
>
>
> Differences (Complex - Receptor - Ligand):
> Energy Component Average Std. Dev. Std. Err. of
> Mean
>
> -------------------------------------------------------------------------------
> VDWAALS -59.5492 2.3688
> 0.3350
> EEL -30.0831 5.7671
> 0.8156
> EPB 50.1463 4.3827
> 0.6198
> ENPOLAR -40.0096 1.0869
> 0.1537
> EDISPER 69.8750 1.3036
> 0.1844
>
> DELTA G gas -89.6322 6.2040
> 0.8774
> DELTA G solv 80.0117 4.7020
> 0.6650
>
> DELTA TOTAL -9.6205 4.1658
> 0.5891
>
> [snip]
>
>
> # DELTA
> # -----------------------
> # MEAN STD
> # =======================
> ELE -30.08 5.83
> VDW -59.55 2.39
> INT 0.00 0.00
> GAS -89.63 6.27
> PBSUR -7.65 0.18
> PBCAL 50.22 4.44
> PBSOL 42.57 4.40
> PBELE 20.13 3.79
> PBTOT -47.07 3.83
> GBSUR -7.65 0.18
> GB 44.45 4.00
> GBSOL 36.80 3.96
> GBELE 14.36 3.19
> GBTOT -52.84 3.16
>
These results are actually not quite as different as you'd think. First of
all, mm_pbsa.pl reports the total electrostatic energy as the sum of the
electrostatic and the 1-4 electrostatic energies. MMPBSA.py, on the other
hand, reports electrostatic energies as ONLY the electrostatic energies.
The 1-4 non-bonded interactions are in some sense 'bonded' or 'internal'
interactions in that they will cancel completely if a single trajectory
calculation is done (correctly). Since this is a good way of checking for
vdW and charge consistency between the prmtops, MMPBSA.py handles them
separately and makes sure that they cancel completely when only a single
trajectory is used for binding energy calculations.
The size of the 1-4 EEL term is typically of the correct size relative to
the EEL term to completely explain the discrepancy between mm_pbsa.pl and
MMPBSA.py in the ELE/EEL terms in the complex, receptor, and ligand. More
importantly, the DELTA EEL/ELE term is exactly the same between
mm_pbsa.pland MMPBSA.py (a good sign). The only term that's slightly
different is
PBCAL/EPB, in which the binding difference is 50.15 in MMPBSA.py and 50.22
in mm_pbsa.pl, which is well within the accuracy of the method.
The reason for the apparently huge difference between MMPBSA.py and
mm_pbsa.pl comes entirely from the non-polar solvation term. MMPBSA.py
allows the PB code to compute the non-polar solvation free energy
contribution. The default scheme in PBSA is to decompose the non-polar
part into a repulsive surface area-based term and an attractive
dispersion-based term (ENPOLAR and EDISPER). The non-polar solvation free
energy in MMPBSA.py computed using this scheme is ~29 kcal/mol. mm_pbsa.pl,
on the other hand, uses the ENP = a*SASA + b form for the non-polar energy
(which is why PBSUR and GBSUR are identical). Therefore, the non-polar
term, which is 29 kcal/mol in MMPBSA.py, is -7.65 kcal/mol in mm_pbsa.pl,
yielding a ~37 kcal/mol difference in the total binding free energy JUST
from the non-polar contribution!
And this accounts for basically the full difference between mm_pbsa.pl and
MMPBSA.py. To my knowledge, there are no plans to expand mm_pbsa.pl to
support the non-polar solvation schemes implemented in PBSA.
FWIW, this non-polar solvation term is responsible for almost all of the
variation found from version to version of PBSA. The electrostatic
solvation free energy fluctuates within 1 kcal/mol or so (as seen between
mm_pbsa.pl and MMPBSA.py).
Hope this helps,
Jason
--
Jason M. Swails
Quantum Theory Project,
University of Florida
Ph.D. Candidate
352-392-4032
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Fri Jun 28 2013 - 07:30:02 PDT
