Re: FEP Rattle Error

From: Ebru Çetin (ebrucetin.ku_at_gmail.com)
Date: Mon Nov 12 2018 - 13:30:08 CST

Hi Brian,

Yes, I have scanned through multiple alchLamba values and they completed
normally. I have tried to tinker with the parameters that are usually
mentioned with the solution of Rattle problem. But, thanks for the caution,
I will try to refrain from changing rigidbonds or timestep.

I am starting from the structure I get from pre-equilibration. I have 200ns
long conventional MD simulation and I am getting different time points
after 50 ns which I minimize and pre-equilibrate for FEP simulation. About
the "over-minimizing" you're right I have done a bit more than necessary
but usually the structures are not getting desired level before 10000
steps. (I know it's a bit huge). I also tried to minimize at lambda = 0 but
it didn't work as well.

Regards,

Ebru

Brian Radak <brian.radak_at_gmail.com>, 12 Kas 2018 Pzt, 21:11 tarihinde şunu
yazdı:

> I think we need a little bit more information.
>
> You said you have already run a full FEP simulation (presumably scanning
> through multiple alchLambda values with runFEP?) and that these completed
> normally. If that is true, you almost certainly should not have to change
> any of the settings that you mention (indeed, changing rigidbonds and/or
> the timestep would change the answer for the free energy).
>
> What structure are you restarting from? FEP calculations are usually very
> sensitive to the initial structure and how representative it is of the
> target alchLambda ensemble. This is why the sequential lambda approach is
> often desirable. However, minimizing and re-thermalizing from lambda = 0 or
> 1 to some other lambda in the middle should be possible and stable.
>
> There is a tendency for NAMD users to "over-minimize" their structures.
> Because of PBCs and how wrapping works in NAMD *this is not a good idea*.
> If you need more than ~200 steps to get a suitable structure for dynamics,
> then there is probably something wrong with your initial structure and you
> are only wasting time. Slow heating is another option, but I've rarely
> since cases where this is truly necessary. In my experience, starting over
> at a temperature of 0 is often more unstable than something closer to 300 K.
>
> HTH,
> BKR
>
> On Sun, Nov 11, 2018 at 6:09 PM Ebru Çetin <ebrucetin.ku_at_gmail.com> wrote:
>
>> Hi,
>>
>> I am trying to rerun some windows on FEP calculations. I have done the
>> simulations from the beginning without a problem, and now I am trying to
>> repeat some problematic windows.
>>
>> In some cases, they give an error due to RATTLE Algorithm. I have tried
>> increasing the box size, increasing margin, disabling wrapAll, and changing
>> rigidbonds to water and time step to 1 fs. When I do change rigidbonds and
>> timestep together error changes to atoms moving too fast.
>>
>> Also, I minimized the system that I get after the equilibration of the
>> first run. I have gone up till 1000 steps since I had minimized it at the
>> very beginning 40000. But it didn't solve the problem as well.
>>
>> I will really be grateful for your help to workaround this problem.
>>
>> Regards,
>>
>> Ebru
>>
>

This archive was generated by hypermail 2.1.6 : Wed Dec 11 2019 - 23:20:23 CST