search

shirtsgroup / cg_openmm

Tools to build coarse grained models and perform simulations with OpenMM
MIT License
20 stars 8 forks source link

Output Data from Replica_exchange.py #4

Open covo6734 opened 4 years ago

covo6734 commented 4 years ago

The data produced by examples/replica_exchange/replica_exchange.py may be repeating instead of generating new data as the parameters are varied. The constants for bond_angle_force_constant and torsion_force_constant have been brought from 0.5 all the day down to 0.001 with no noticeable change on the output .pdb files. Additionally, the graphical output for the Replica Exchange Simulation shows nearly identical results for the energy values regardless of what these constant values are set to. This seems to indicate the simulation is not accounting for these constants, or resetting them before each run. Changing the temperature ranges also seems to have little to no effect on these outputs.

mrshirts commented 4 years ago

What version of yank are you using? I'm having issues running it with Yank 0.20.1?

covo6734 commented 4 years ago

The yank on the computer I've been running from is Yank 0.21.2

mrshirts commented 4 years ago

So, I have updated Yank to 0.25.2, and am getting a No module named 'yank.multistate' error. Since Yank is getting updated, my current plan is to debug in Yank 0.25.2, and then have other people update yank. Does that make sense?

mrshirts commented 4 years ago

OK, I have the example running. I've posted the changes I needed to make with up-to-date libraries in branch mrs_tweak_feb2020. But very strange that the energy is so low for all attempts - seems to be of order `10^-19 (are energy off by a factor of N_avogadro), and the positions are moving quite slowly for that number of steps I'll keep poking around with this - perhaps a unit conversion error hidden somewhere?

covo6734 commented 4 years ago

I checked and did not find any unit conversion errors. It does appear that energy only varies with temperature and not the torsion_force_constant or the bond_angle_force_constant. The output of the rep_ex_ener.png graph only changes in regards to temperature.

mrshirts commented 4 years ago

Interesting, so the model may not be including those things. I'll keep poking.