Convert Lipid21 to FFXML

[gohdavid]

I wanted sphingomyelin parameters from Amber Lipid21 and couldn't find an ffxml online. I tried converting it using convert_amber.py, but ParmEd returned an error due to having two different scnbs:

NotImplementedError: Cannot currently handle mixed 1-4 scaling: 1-4 eel [1.2] 1-4 vdw [2.0, 6.0]

From $AMBERHOME/dat/leap/parm/Lipid21.dat:

cD-cD-cD-cD    1    0.100    0.000  -5.0    SCEE=1.2 SCNB=6.0  Lipid21 v1.0 MP2 QM fit.
cD-cD-cD-cD    1    0.100    0.000  -4.0    SCEE=1.2 SCNB=6.0  Lipid21 v1.0 MP2 QM fit.
cD-cD-cD-cD    1    0.300    0.000  -3.0    SCEE=1.2 SCNB=6.0  Lipid21 v1.0 MP2 QM fit.
cD-cD-cD-cD    1    0.100    0.000  -2.0    SCEE=1.2 SCNB=6.0  Lipid21 v1.0 MP2 QM fit.
cD-cD-cD-cD    1    0.100    0.000   1.0    SCEE=1.2 SCNB=6.0  Lipid21 v1.0 MP2 QM fit.
cA-oT-pA-oP    1    0.000    0.000  -5.0    SCEE=1.2 SCNB=2.0  Lipid21 v1.0
cA-oT-pA-oP    1    0.000    0.000  -4.0    SCEE=1.2 SCNB=2.0  Lipid21 v1.0
cA-oT-pA-oP    1    0.000    0.000  -3.0    SCEE=1.2 SCNB=2.0  Lipid21 v1.0
cA-oT-pA-oP    1    0.200    0.000  -2.0    SCEE=1.2 SCNB=2.0  Lipid21 v1.0

Looking through #156 and ParmEd/ParmEd#1136 it seems like there was a fix for this when there's only one non-unity scnb.

I think it would be neat to implement Lipid21 in OpenMM and was wondering what needs to be done to implement two 1-4 scalings in OpenMM.

Or, in case anyone knows of an ffxml for Lipid21, could you share it with me, please?

[jchodera]

@peastman : Does the openmm.app.ForceField pipeline permit a way to have different .ffxml files specify distinct 1-4 scalings? If so, there's likely a way for us to generate a separate .ffxml file for Amber Lipid21 that specifies the distinct scaling factors.

[jchodera]

@gohdavid : In the meantime, you can still set up your system in AmberTools LEaP and read it into OpenMM via AmberPrmtopFile!

[peastman]

It requires using a script to postprocess the exceptions. See the discussion and sample code in ParmEd/ParmEd#1136 and ParmEd/ParmEd#1149