Minor issue regarding residue names.

This is a relatively minor issue, but the name of the residue seems to affect the free energy calculation in unpredictable ways. In example1, all solvent molecules are labeled as 0C9 in the .yaml file and in SYSTEM.prmtop. The simulation runs normally but yields the strange result of 67.851 ± 0.128 kcal/mol.

example1.zip

In example2, all solvent molecules are labeled as XXX in the .yaml file and in SYSTEM.prmtop. A simulation run with these files yields the correct result of -5.140 ± 0.094 kcal/mol. The reference value for this solvation free energy is -5.2 ± 0.1 kcal/mol.

example2.zip

I think it would be interesting to either interrupt a simulation in the beginning if the residue name starts with a number or make Yank accept these unusual names. You can find more details in the attached files.

Thanks!

[andrrizzi]

This looks like an MDTraj problem. With

import mdtraj

topology = mdtraj.load_topology('SYSTEM.prmtop')
print('N atoms: ', topology.n_atoms)
print('N residues', topology.n_residues)
print('N ligand atoms: ', len(topology.select('resname LIG')))
print('N solvent atoms: ', len(topology.select('resname 0C9')))

I got

N atoms:  2822
N residues 157
N ligand atoms:  14
N solvent atoms:  0

I’ll open an issue there.

[andrrizzi]

I think I implemented a work-around in #819 that was released with YANK 0.19 to avoid Leap crashing when a number was in the residue name. If you’re using a version >= 0.19 that might be the culprit.