Re: [AMBER] EGB and ESCF might not be converged if the net charge not equal to 0 in the QM/MM-GBSA

From: Ross Walker <ross.rosswalker.co.uk>
Date: Mon, 6 Apr 2015 09:21:03 -0700

Hi Henry,

My question was not related to this and you appear to overanalyzing at this point. My question was are your QM calculations reasonable? Have you taken a careful look at the QM output to make sure the simulations are actually converging?

All these wild jumps to me look like the QM calculation is probably misbehaved. Very possible given the SCF has been modified in later version of AMBER since I originally wrote it back in Amber 9. I'd suggest starting there. Check for things like charge neutralization messages in the QM calculations - make sure the code isn't forcing some kind of neutralization etc. If you see issues with the QM calculation this is something it might be worth trying with an earlier version of AMBER, like AMBER 9 or AMBER 10 - those versions were much much better tested for QM/MM and will likely be faster and I suspect suffer from less convergence issues.

Start by breaking out the QM calculations that show the jumps and look at the output carefully. If you can provide me with the input files necessary for a simple test case that shows the issue you are having I can take a look, and or try it with Amber 9 and see what happens.

The key is to rule out problems with the simulations themselves before going to far down the rabbit hole looking for a complex charge transfer / other physical explanation.

All the best
Ross


> On Apr 5, 2015, at 5:35 PM, psu4.uic.edu wrote:
>
> Dear Professor Walker,
>
> The problem here is once NADH (the net charge=-2) is included in the
> QM/MM-GBSA calculation (as shown in the *Figure c & d *here
> <https://drive.google.com/file/d/0B84zOE6DopPlLU9BVTd1SDk0aUE/view?usp=sharing>)
> , the dGsolv (blue) and dGscf (green) show extreme instantaneous values
> (dGscf<green> in dGbind can fluctuate really widely: -700~1000 kcal/mol
> for example ). Therefore the abnormal behavior makes large and positive
> dGbind (for example, dGbind = 34.37 kcal/mol. The dGbind calculated from
> MM/GBSA using the same trajectories here is -30 kcal/mol). This happens in
> igb=1,2,5,7,8 and all effective radii settings. On the other hand, if only
> the ligand (net charge=0) or the ligand +Tyr156 (net charge=0) are assigned
> as QM regions, the QM/MM-GBSA calculations are normal with converged
> dGscf *(Figure
> a & b)*.
>
> Professor Monard suggests that "If the MM and the QM regions are charged,
> the main component of the electrostatic QM/MM interaction is a
> charge-charge interaction
> whose value is the total charge of the QM region times the total charge of
> the MM region divided by the distance between the two centers. If you are
> "lucky", these two centers are close and the electrostatic interactions is
> very large." Therefore, in order to avoid the charge-charge interaction,
> we tried to systematically reduce the "qmcut" from the default 9,999A to 8
> & 0 A, as shown in* Figure d-f*. However, the extremely positive and
> negative dGscf (green lines) still happened, though less frequently.
>
> The professor Monard further enlightens that "if you have 1 QM region
> in the receptor (e.g.,NADH), and the ligand is QM, when you form the
> complex, you have one big QM region where all molecular orbitals will be
> delocalized between NADH
> and the ligand. This could mean huge charge transfer and polarization
> components."
>
> In response, we run QM/MM-GBSA where the "ligand" is NADH+ligand
> (qmcharge_lig=-2, qmcharge_com=-2, qmcharge_rec=0), the extreme values of
> electrostatic interactions energy terms didn't happen with qmcutt=0, 8, and
> 9999 as shown in the *figures g-i*. However, the dGbind here is
> unrealistic since the "NADH+ligand" is deemed as the "ligand". However,
> this result seems to confirm the saying that the huge charge transfer and
> polarization components between NADH and the ligand when the system forms
> the complex.
>
> I guess the charge settings in the above file should be correct? Could
> you kindly let us know if you see any setting mistakes? If not, it seems
> like including the charged NADH as part of the QM receptor and QM complex
> region in the QM/MM-GBSA calculation is the cause of the extreme dGscf
> values due to large charge-charge interactions between non-neutral QM and
> MM regions AND huge charge transfer and polarization components between the
> QM ligand and NADH regions.
>
> Best,
> Henry
>
> On Fri, Apr 3, 2015 at 5:57 PM, Ross Walker <ross.rosswalker.co.uk> wrote:
>
>>>
>>> The conclusion here seems to go back to the original point: if both QM
>>> and MM regions are non-neutral, the QM/MM calculations are problematic.
>> If
>>> the QM region includes NADH (net charge= -2) and the ligand, reducing the
>>> qmcut even to 0 cannot solve the issue in our case. The huge charge
>>> transfer and polarization components between NADH and the ligand, during
>>> the complex forming stage, still creates extreme electrostatic values.
>>
>> Henry, what is your definition of being problematic - is the SCF not
>> converging? - which means anything calculated from that is incorrect. Or is
>> the QM calculation working but the ultimate GBSA binding result is coming
>> out incorrect?
>>
>> The QM code should support running simulations with non zero charges
>> (although the QM region requires an integer number of electrons). Are you
>> setting the charge correctly in the QM calculation? - I could see how
>> enforcing the charge to be zero but including 2 extra electrons could cause
>> all sorts of problems.
>>
>> All the best
>> Ross
>>
>>
>>
>>
>> /\
>> \/
>> |\oss Walker
>>
>> ---------------------------------------------------------
>> | Associate Research Professor |
>> | San Diego Supercomputer Center |
>> | Adjunct Associate Professor |
>> | Dept. of Chemistry and Biochemistry |
>> | University of California San Diego |
>> | NVIDIA Fellow |
>> | http://www.rosswalker.co.uk | http://www.wmd-lab.org |
>> | Tel: +1 858 822 0854 | EMail:- ross.rosswalker.co.uk |
>> ---------------------------------------------------------
>>
>> 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.ambermd.org
>> http://lists.ambermd.org/mailman/listinfo/amber
>>
>
>
>
> --
>
> Pin-Chih Su (Henry Su)
>
> Ph.D. canditate
>
> Center for Pharmaceutical Biotechnology (MC 870)
>
> College of Pharmacy, University of Illinois at Chicago
>
> 900 South Ashland Avenue, Room 1052
>
> Chicago, IL 60607-7173
>
> office 312-996-5388
>
> fax 312-413-9303
> _______________________________________________
> 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 Mon Apr 06 2015 - 09:30:02 PDT
Custom Search