Making HMR work with dummies about biosimspace HOT 11 CLOSED

Yes, I agree. I'll check how things are implemented in SOMD and see if @jmichel80 has any input. It's an easy fix to make once we agree on the correct approach. (It's almost like you want to do the partitioning before the merge, i.e. just add the dummies afterwards.)

from biosimspace.

Thanks for this. I hadn't thought about dummies when writing the code and agree with your suggested solution. I'll check what SOMD does too, since I believe I largely followed the implementation there, or at least tested mine against that one when first written.

from biosimspace.

I think the issue is that the repartitioning code generates a connectivity object to working the bonding when performing the HMR. This doesn't work properly if an atom is a dummy, since it's based on equilibrium bond radii for specific elements. I think I'll need to adjust this to use the elements from the non-dummy end state instead. Hopefully this should be easy enough to fix.

Thanks for reporting.

from biosimspace.

Thanks @lohedges, would this work if both endstates have dummies? I think that's the common use case.

from biosimspace.

Yes, assuming the velocity conversion is okay I think capping with a warning would be an okay solution.

from biosimspace.

I think there are two approaches:

• Include dummies in the repartitioning by using the element property from the opposite end state. In this case dummies are part of the repartitioning so the mass of the end state is consistent before and after repartitioning, but the mass of a dummy atom isn't necessarily the same between end states as it was before.
• Exclude dummies from the repartitioning, then set their mass to that of the opposite end state after repartitioning. In this case, the overall molecule mass of the end state may differ before / after repartitioning (if the sum includes dummies) but the dummy mass is consistent between end states.

from biosimspace.

Yes I was thinking about that as well, it seems to me that the second approach makes more sense, as one would not expect conservation of mass in an alchemical perturbation setting anyway as you are adding dummies regardless.

from biosimspace.

SOMD uses the second approach above, i.e. excluding dummy atoms and replacing with the repartitioned mass at the other end state. I'll implement this and check that things are working as expected.

from biosimspace.

Okay, this is now fixed, i.e. I get the same behaviour as SOMD, which agrees with the approach agreed upon above. This requires a new build of Sire, so I'll submit a PR when that's done. (Likely Monday.)

Thanks again for catching this.

from biosimspace.

Many thanks for that @lohedges, I am off until Jan 3rd, so I will only be able to test it then. Happy holidays!

from biosimspace.

Thanks, you too. Have a good break!

from biosimspace.