HowTo

-----------------------------------------------------------------------------
FFAFFURR -- Framework For Adjusting Force Fields Using Regularized Regression
-----------------------------------------------------------------------------

Because the goal is to "adjust" existing parameters of the OPLS-AA FF, it is reasonable to provide a standardized listing of parameters as input.
This is achieved for any system already set-up with TINKER, e.g. when using the standard OPLS-AA FF parameters that are distributed with the TINKER package (in this work version 7.1.2 of the TINKER package was used), by using the analyze tool, i.e. by issuing the following command:

analyze -k sample.key sample.xyz P ALL > ffaffurr.input.originalFF

The sample.key file thereby denotes the keyword parameter file in TINKER that most importantly contains the location of the potential energy parameter file, e.g. oplsaa.prm for the standard OPLS-AA FF distributed with the TINKER package.
The sample.xyz file is the basic TINKER coordinate file type.
The standardized listing of parameters is redirected into the ffaffurr.input.originalFF file that needs to be provided as input to FFAFFURR by placing it in the same directory as the Python file ffaffurr.py.
In a similar fashion, a standardized connectivity list for each of the atoms may be generated by issuing the following command:

analyze -k sample.key sample.xyz C ALL > ffaffurr.input.interactionsFF

As before, the ffaffurr.input.interactionsFF file needs to be provided as input by placing it in the same directory as the Python file ffaffurr.py.

Third, the input file ffaffurr.input.FHI-aims-logfiles contains a list of FHI-aims-specific output files produced when calculating single-point DFT energies.
Obviously, these files must be produced for a set of conformers that serves as training data.

Finally, the input file ffaffurr.input contains the "switches" that control the behavior of the framework, e.g. what kind of parameters are to be "adjusted" or what regression model to use.