You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
It seems that the energy spiking issue #86 may be due to an issue with the angle/torsion nonbonded exclusions. It seems that some nonbonded interactions are not being enforced. For a single Langevin dynamics simulation at 400K with LJ epsilon=5kJ/mol, and bond angle potentials turned on, a pairwise rdf distribution for backbone-sidechain distances shows that bb and sc beads are frequently right on top of eachother. Naturally, this makes the system very unstable.
Torsions with a force constant of 0 should be ok, provided that the 1-4 nonbond scaling is set 1 (which it is currently unless we are considering rosetta exceptions).
We should allow for certain angle types to keep their 1-3 nonbonded interactions, rather than setting a default force constant which is applied to all angles. Or in the short term, make sure that a force constant of zero is not allowed.
However I think the spiking issue is more than this - here's what we see for different variations:
include_bond_angle_forces=False, include_torsion_forces=False --> no spiking
include_bond_angle_forces=True, all k_angle are 25 kJ/mol/rad^2, include_torsion_forces=True --> spiking
include_bond_angle_forces=True, all k_angle are 25 kJ/mol/rad^2, include_torsion_forces=False --> spiking
It seems that the energy spiking issue #86 may be due to an issue with the angle/torsion nonbonded exclusions. It seems that some nonbonded interactions are not being enforced. For a single Langevin dynamics simulation at 400K with LJ epsilon=5kJ/mol, and bond angle potentials turned on, a pairwise rdf distribution for backbone-sidechain distances shows that bb and sc beads are frequently right on top of eachother. Naturally, this makes the system very unstable.
Torsions with a force constant of 0 should be ok, provided that the 1-4 nonbond scaling is set 1 (which it is currently unless we are considering rosetta exceptions).
We should allow for certain angle types to keep their 1-3 nonbonded interactions, rather than setting a default force constant which is applied to all angles. Or in the short term, make sure that a force constant of zero is not allowed.
However I think the spiking issue is more than this - here's what we see for different variations:
The text was updated successfully, but these errors were encountered: