Discovery of nitrogen fixation catalysts with genetic algorithms

Code repository for the paper Discovery of molybdenum based nitrogen fixation catalysts with genetic algorithms.

1. Run
2. Parameters

How to run

To setup an environment for running the GA, activate a conda environment with pip installed and run:

pip install .

Then ensure you have installed xtb in the environment.

To do a simple GA run, activate the environment and run:

python genetic_algorithm_driver.py --scoring_func calculate_score --partition kemi1 --generations 1 --filename data/ligands.smi --debug --population_size 4 --molecules_per_chunk 2 --cpus_per_chunk 2 --sa_screening --CovalentLigands 3 --DativeLigands 1 --opt loose --n_confs 2

Input driver arguments

A list of possible arguments and their discriptions are seen below:

Arg	Description
-h or --help	Prints help message.
--population_size	Sets the size of the population pool.
--n_confs	Sets how many conformers to generate for each molecule.
--num_rotatable_bonds	Set the limit for the number of rotatable bonds for new children.
--max_size	Maximum number of heavy atoms allowed in crossover
--min_size	Minimum number of heavy atoms allowed in crossover
--DativeLigands	Number of dative ligands in each catalyst
--CovalentLigands	Number of covalent ligands in each catalyst
--BidentateLigands	Number of bidentate ligands in each catalyst
--molecules_per_chunk	Parallelization option. How many molecules to put in each SLURM job
--cpus_per_chunk	Parallelization option. How many cpus to reserve for each SLURM job
--memory_per_chunk	Parallelization option. How much memory per cpu in a SLURM job
--partition	Cluster to run on
--debug	If set the starting population is a set of 4 small molecules that can run fast locally. Used for debugging.
--find_connection_point	Pre-screen ligand candidates for connection points
--load_from_pickle	Load starting population from pickle file
--minimize_score	Minimize the score obtained from scoring function
--generations	How many evolution cycles of the population is performed.
--n_tries	Sets how many times the GA is restarted. Can be used to run multiple GA runs in a single submission.
--scoring_function	Chose one of the specified scoring functions to run
--sa_screening	Decides if synthetic accessibility score is enabled. Highly recommended to turn this on.
--filename	Path to the database extract to create starting population.
--output_dir	Sets output directory for all files generated during generations.
--scratch	Path to put any intermediate scratch files
--cleanup	If enabled, all scoring files are removed after scoring. Only the optimized structures and their energies are saved.
--method	Which gfn method to use.
--timeout_min	How many minutes each slurm job is allowed to run
--rms_prune	Pruning value for RDKit conformer embedding
--energy_cutoff	Sets energy cutoff on the conformer filtering.
--opt	Set optimization convergence criteria for xTB.

Authors

Magnus Strandgaard¹

¹ Department of Chemistry, University of Copenhagen, 2100 Copenhagen Ø, Denmark. Email: [email protected].