Raw Data for "Data-Driven Modeling of Hoveyda–Grubbs-II-Type Olefin Metathesis Precatalyst Initiation: Mechanistic Insights and Predictive Models"

Additionally, we integrated multiple scattered structure-reactivity relationships from previous studies into a coherent model of precatalyst activation, including the trans-influence of the L-type carbene ligand, and the steric effect of the O-substituent, highlighting that Ru–O bond strength is only one aspect among many. These findings indicate that stabilizing key steps like ether dissociation and olefin coordination offers a promising alternative strategy for designing fast-initiating precatalysts, complementing the established method of precatalyst destabilization.

Finally, we trained a machine learning model ensemble to predict activation barriers for 122 new precatalysts. DATASET DESCRIPTION: The DFT-optimized geometries of all precatalysts and used intermediates "Int1" (as named in the manuscript) are given in the 'Geometries' directory in the XYZ format. The directory structure differentiates between the precatalyst and the intermediate Int1, followed by a seperation of precatalysts obtained from the data mining and precatalysts used for the predictions with the ML (machine learning) ensemble.

The numbering of the .xyz-files follows the one used in the manuscript and supporting information. In the 'Notebook' directory, all data needed to reproduce the MLR (multiple linear regression) and ML models are given. The data is given as .csv (readable with any text editor). - dG_exp.csv contains the data mined initiation rate constants converted to the activation barriers which are used to train the MLR and ML models - ML_descriptors.csv contains the descriptors used to train the ML ensemble models - ML_results.csv contains the predictions of each of the ML ensemble models - MLR_descriptors.csv contains all descriptors used in the MLR models presented in the manuscript - SMILES.csv contains all SMILES strings for all precatalysts in this work -- Notebook_Models.ipynb is a jupyter notebook containing all python code needed to conveniently reproduce all MLR and ML models together with the data given in the .csv-files.

To execute the notebook, Python3 needs to be installed together with the following packages: numpy, pandas, matplotlib, scikit-learn, RDKit