mmunibas / kmdcm

Kernel-based Minimal Distributed Charge Models (kMDCM)

Installation

After cloning the repository, you can install the package by running the following commands:

python3 -m venv kmdcm-env
source kmdcm-env/bin/activate
pip install -r requirements.txt
pip install -e .

PyDCM requires the Fortran code to be compiled. To do so, run the following commands:

export CFLAGS="-std=c99"
cd kmdcm/pydcm
bash compile_f2py.sh

Usage

First, create the input json file. An example is given in "create_json_input.py". To create the json file, run the following command:

python kmdcm/pydcm/create_json.py

To optimize the charges and fit the kernels, run the following command in the "tests" directory:

Example 1: Water

 python test_dcm.py --alpha 0.000001 --n_factor 8 --n_atoms 3  --l2 0.0 --json water_pbe0.json --fname water_pbe0 --do_opt

Example 2: Methanol

python test_dcm.py --alpha 0.000001 --n_factor 16 --n_atoms 6  --l2 40.0 --json shaked-methanol.json --fname shaked-methanol --do_opt

Keyword arguments

• --alpha is the regularization parameter
• --n_factor is the ratio between the test and training set
• --n_atoms is the number of atoms in the molecule
• --l2 is the l2 penalty for moving the charges from their initial positions
• --json is the json file containing the filenames for fitting
• --fname is the name of the job

Exporting kernel matrices to CHARMM

To export the kernel matrices to CHARMM, run the following command:

python kmdcm/utils/save_charmm_input.py --kernel 4af4d6a1-a66c-4339-bafa-82db5e7529fc

This creates a directory in the tests/coeffs which stores the kernel coefficients ("coefs0.txt", etc.) and the reference distances ("x_fit.txt").

156 10 3 6
acec8c45-bfe5-4acc-8574-925372ecb40d/x_fit.txt
acec8c45-bfe5-4acc-8574-925372ecb40d/coefs0.txt
...
acec8c45-bfe5-4acc-8574-925372ecb40d/coefs29.txt

Running kMDCM in CHARMM

! 
open unit 12 card read name @input/methanol.kern
! regular MDCM input file
open unit 10 card read name @input/10charges.dcm
DCM KERN 12 IUDCM 10 TSHIFT 
'''