QM Optimisation fails

[dombrowsky]

Hi,

I am currently trying to generate parameters for a molecule and it consistently fails in the optimization step because no conformer was found that could be optimized to GAU TIGHT. I am running qubekit with psi4 using 6-311++G as basis and wb97x-d as method. Charges are handled with MBISCharges in psi4 with water as medium solvent and angstrom as units. Additionally, I added the Rfree parameters for sulfur from the Paper. Parametrisation is handled by OpenFF.

I have tried to up the convergece criteria of the pre-optimization run to GAU VERYTIGHT but it did not change the outcome. I also tried to ramp up the number of maximum iterations to 500 and the number of seed conformers to 60 but again this did not help.

Maybe there are some special configurations that I am not aware of that you could hint me to?

The molecule is: "C=c5cc(N1CCOCC1)oc6c(c3cccc4c2ccccc2sc34)cccc56"

[jthorton]

Hi, @dombrowsky thanks for the report it seems like the pre-opt method is not helping as much as we hoped in this case. I would recommend trying to pre-optimise the molecule using QM at some cheaper theory level which might get you close to the minimum either using qubekit or separately and use that as input into the workflow.

I am running qubekit with psi4 using 6-311++G as basis and wb97x-d as method. Charges are handled with MBISCharges in psi4 with water as medium solvent and angstrom as units. Additionally, I added the Rfree parameters for sulfur from the Paper.

Note that the Rfree parameters were not optimised for this combination of method, basis and implicit solvent so their performance will not be optimal.

[dombrowsky]

Hi, @jthorton thank you for your reply. I will try to follow your advise and try another theory level.

On another note, can you by any chance point me to a reference for Rfree parameters that work with psi4? Currently with psi4, only MBISCharges seem to work. Is there another basis and method that the parameters might work better with or am I out of luck there?

Best regards,

Max Dombrowsky

[jthorton]

@dombrowsky our most recent publication details all of the combinations of QM method, basis, implicit solvent, charge partitioning method etc that we have parameterised the Rfree for. These can all be accessed from qubekit as protocols. For example, running your molecule with protocol 5e which works with MBIS and B3LYP-D3BJ/DZVP can be done with the following command qubekit run -sm "C=c5cc(N1CCOCC1)oc6c(c3cccc4c2ccccc2sc34)cccc56" -n "mol1" -p 5e.

If you would like to make a config file for a protocol and view the parameters you can use
qubekit config create protocol_5e.json -p 5e

I hope this helps!