Hi David,
I have a follow up question to this topic, I know now the ligand I am looking at switches between two possible conformation, one gives higher MMPBSA and the other one lower values. It’s clear from the experimental data that the correct values matches the lowest mmpbsa values conformation.
But system continues to switch between conformations, even after 2 micro seconds. So I wonder how to deal with that, as I am indeed looking for convergence, so I can believe the values when I add for example a similar ligand or additional mutants. Averaging does not seems a good strategy, maybe heating?
Any suggestions will be greatly appreciated,
Fabian
-------------------
Fabian Glaser
Bioinformatics Knowledge Unit - BKU
The Lorry I. Lokey Center for Life Sciences and Engineering
Technion - Israel Institute of Technology, Haifa, Israel
https://lokey.technion.ac.il/bioinformatic-knowledge-unit/
Tel +972 (0) 4 8293701
> On 5 Dec 2021, at 16:04, David A Case <david.case.rutgers.edu> wrote:
>
> On Sat, Dec 04, 2021, Fabian Glaser wrote:
>>
>> I am working on agonist ranking project, computing free energy for wt and
>> mutants. I tried several MMPBSA input values, and it seems that inp=1 is
>> better than inp=2. inp=1 values give values more similar to experimental
>> values, at least for some examples.
>
> My experience is limited and quite a bit out of date, but here goes: the
> inp=2 option seems a bit more "fragile" than does the simpler inp=1 option,
> especially for absolute binding energies. I too ran into cases where inp=2
> gave very odd results, but the details are lost in the mists of time. Maybe
> Ray and folks from his group can help here.
>
>> I am running multiple trajectories, and I found the delta total values are
>> very variable with a stdev of about +-3-5, specially for the mutants. I
>> tried running very long simulations, but still the values drift a lot. I
>> also tried different sampling using 500 or 1000 frames, but it did not
>> improve the results.
>
> Answers here are surely system-specific. Try using more frames, and/or
> different length simulations. Note that you are looking for convergnence
> here, not for agreement with experiment.
>
> Be sure that your system is stable, and not always exploring new
> configurations as time progresses. Check results for (say) the first half
> of a given trajectory vs. the second half.
>
> ...good luck...dac
>
>
> _______________________________________________
> 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 Wed Dec 08 2021 - 06:00:03 PST
