PE is NaN error during ASFE runs #1259
-
|
When running absolute solvation free energy, all the vacuum legs finish fine, but some of the solvent legs sporadically throw this error:
Any advice on what to do when running into this? |
Beta Was this translation helpful? Give feedback.
Replies: 1 comment 5 replies
-
|
Hi @CurtisColwell ! This is rather interesting. By any chance is your ligand quite big? I have noticed this issue in cases where there's a large change in box volume during the equilibration (we've got some kind of fix in one of our branches), but it's not a bug we've been able to consistently reproduce. |
Beta Was this translation helpful? Give feedback.
-
|
Not crazy big. MW is 380 and HAC 27. On the small side for a typical drug. It is definitely not an easy or cheap thing to test. It automatically restarts the run and always eventually finishes, it just wastes GPU time. I'll try increasing the padding and see if that helps. Thanks! |
Beta Was this translation helpful? Give feedback.
-
|
I increased the box size to 1.5 nm and after one run do not see this error. I'll keep you posted. |
Beta Was this translation helpful? Give feedback.
-
|
To follow up, the occurrence of the NaN error is definitely proportional to the box size during the equilibration. Could you point me to the fix? I would like to try it. I'm starting at 1.5 nm and increasing with 2 nm if it errors out, but this is getting slow and expensive. |
Beta Was this translation helpful? Give feedback.
-
|
@CurtisColwell the minimal fix can be found in this pull request: #1275, If you use it alongside the latest openmmtools release, you should get more stable stimulations. |
Beta Was this translation helpful? Give feedback.
-
|
Thanks! I will test this and let you know how it goes. |
Beta Was this translation helpful? Give feedback.
@CurtisColwell the minimal fix can be found in this pull request: #1275,
If you use it alongside the latest openmmtools release, you should get more stable stimulations.