/NiCOlit

Repository for the featurization of the NiCOlit reaction dataset and machine learning model training for yield prediction

Primary LanguageJupyter NotebookMIT LicenseMIT

Prediction of reaction yields on a chemical diversity calibrated dataset: the NiCOLit dataset

This is the code used to generate the results published in Machine Learning Yield Prediction from NiCOlit, a Small-Size Literature Data Set of Nickel Catalyzed C–O Couplings

A preprint version is available on ChemArxiv.

The NiCOlit dataset is accessible here.

Paris

Description :

A description of the different folders is given below :

  • aqc_utils : this folder contains usefull tools released by Auto-QChem.
  • data :
    • the NiCOlit dataset downloadable here.
    • HTE : HTE datasets published by D. T. Ahneman et al. and A. B. Santanilla et al. and used in a publication of P. Scwhaller.
    • utils : csv files of DFT molecular featurization needed for DFT-featurization.
    • rxnfp_featurization : csv files of Ahneman, Santanilla and NiCOlit datasets featurized with the RXNFP method.
  • descriptors :
    • preprocessing of the NiCOlit dataset.
    • DFT and RDKit featurisation.
  • images : All images displayed in the article.
  • notebooks_dft : All notebooks allowing to compute the DFT featurizations of molecules in the dataset
  • notebooks_ord : All notebooks necessary for producing the csv for ORD submission of NiCOlit
  • notebooks_dataset_analysis:
    • 01_visualization_chemical_analysis: Displays the relative occurences of coupling-partner/substrate/ligand pairs in NiCOlit (Figure 1 of the paper)
    • 02_visualization_diversity_and_scope_optimization_structure: Analysis of the dataset's chemical diversity and scope optimisation structure (Figure 2 and 3 of the paper)
    • 03_visualization_chemical_space_exploration: Displays the dataset with an analysis of DOI/coupling partners/substrate repartition within the dataset (Figure 4 of the paper)
    • 04_visualization_nickel_precursors: visualiszing catalysts precursors
  • notebooks_prediction_results: all notebooks necessary to run the different machine learning models
  • notebooks_prediction_visualisation: all notebooks necessary to analyze the predictions of machine learning models

Install requirements

Best use with following requirements :

pip install -r requirements.txt

Reproducing figures and experiments

In order to reproduce the figures and experimens from the paper, you can run in order the notebooks in the different notebooks sections (except the ones in the notebooks_dft, that were used to compute the DFT descriptors which are direclty accessible in the data folder). The only prerequisite is running the notebooks from notebooks_prediction_results before those from notebooks_prediction_visualization.