Running and analyzing SAMS
See original GitHub issueHi everyone, I’ve been running some simulations using SAMS on the BRD4 dataset, and I have some questions I hope you could answer about using SAMS.
First off, what are your suggested best practice protocols for running SAMS? Would it be similar to what you have for the SAMS sampler in the BRD4 example yaml in the yank-examples repo?
Another question I have pertains to the SAMS sampling. What should I expect to see in the yank anlayze report
output if SAMS is working correctly and should give converged results, or-conversely-when the sampling of SAMS is inadequate?
Finally, I have a question about some strange behavior I’ve been seeing during the analysis of the SAMS simulations I have run (perhaps this ties in with my previous question). I’ve been running some BRD4 simulations for about 2 days now with GPUs, which are still running. From the mixing and replica exchange plots I would think there’s perhaps enough sampling to get converged binding free energy estimates.
However subsequent analysis on the running simulation data spaced a few minutes apart gives completely different estimates, which makes me think that I’m overlooking something important. I’d appreciate your thoughts.
2018-06-18 16:48:58,946: Free energy of binding : -22.698 +- 1.598 kT (-13.532 +- 0.953 kcal/mol)
2018-06-18 16:48:58,947:
2018-06-18 16:48:58,947: DeltaG complex : 187.216 +- 0.948 kT
2018-06-18 16:48:58,947: DeltaG restraint : -3.039 kT
2018-06-18 16:48:58,947: DeltaG solvent : 161.479 +- 1.287 kT
2018-06-18 16:48:58,947:
2018-06-18 16:48:58,948: Enthalpy of binding : 136.134 +- 153.938 kT (81.158 +- 91.772 kcal/mol)
#analysis two minutes later
2018-06-18 16:50:19,786: Free energy of binding : -2.350 +- 1.466 kT (-1.401 +- 0.874 kcal/mol)
2018-06-18 16:50:19,786:
2018-06-18 16:50:19,787: DeltaG complex : 187.216 +- 0.948 kT
2018-06-18 16:50:19,787: DeltaG restraint : -3.039 kT
2018-06-18 16:50:19,787: DeltaG solvent : 181.827 +- 1.117 kT
2018-06-18 16:50:19,787:
2018-06-18 16:50:19,787: Enthalpy of binding : 291.468 +- 153.914 kT (173.762 +- 91.758 kcal/mol)
#analysis 5 minutes after the previous analysis
2018-06-18 16:55:49,895: Free energy of binding : -44.349 +- 1.560 kT (-26.439 +- 0.930 kcal/mol)
2018-06-18 16:55:49,896:
2018-06-18 16:55:49,896: DeltaG complex : 190.218 +- 0.753 kT
2018-06-18 16:55:49,896: DeltaG restraint : -3.039 kT
2018-06-18 16:55:49,896: DeltaG solvent : 142.830 +- 1.366 kT
2018-06-18 16:55:49,896:
2018-06-18 16:55:49,897: Enthalpy of binding : -65.223 +- 413.379 kT (-38.884 +- 246.441 kcal/mol)
I’ve attached a zip file that contains the yaml file used to run the SAMS simulations and the analysis .ipynb in case you have other comments or suggestions for protocol improvements. sams.zip
Thanks for your help!
Issue Analytics
- State:
- Created 5 years ago
- Comments:21 (15 by maintainers)
Top GitHub Comments
@jchodera @Lnaden Sure. I’ll get the simulations with exact PME started up now, and I’ll ping you folks when I get the results back.
@lnaden: Looks like this is an example where the routines to automatically omit “bad” netcdf chunks aren’t working as expected.