Betaine Catalyst Repository

This repository contains the data, descriptor-generation outputs, modeling workflows, and DFT/kinetic analysis files used for the work: "A Predictive Betaine Organocatalyst Platform for Asymmetric 1,3-Dipolar-Cycloadditions Enabled by Interaction-Aware Modeling" (https://doi.org/10.26434/chemrxiv.15006640/v1). It is organized as follows:

Repository Map

Path	What is inside	Start here
dataset.xlsx	Experimental data workbook. comb-matrix contains the main catalyst-substrate matrix; N-substitution contains the N-substitution comparison set.	Open the workbook when checking source experimental values.
betaine_env.yml	Conda environment for the local Python notebooks and helper scripts.	conda env create -f betaine_env.yml
descriptor_generation/	Molecule lists, Gaussian log files, compact descriptor tables, and descriptor-combination scripts.	descriptor_generation/ReadMe
catalyst-substrate-modeling/	Main LASSO modeling notebook, prepared feature CSVs, virtual-screening input, and saved model outputs.	catalyst-substrate-modeling/README.md
catalyst-only-modeling/	Catalyst-only feature workbook and one saved parity plot from the upstream SigmanGroup modeling workflow.	catalyst-only-modeling/README.md
DFT_reaction-mechanism/	Optimized structures, reaction-state energies, transition-state conformer searches, kinetic simulation notebook, and DFT support scripts.	See the folder notes below.

Data And Workflow Overview

1. All experimental outcomes are listed in dataset.xlsx.
2. Molecular descriptors are prepared under descriptor_generation/.
3. Combined descriptor matrices are used by the modeling folders:
   • catalyst-substrate-modeling/ for the main catalyst-substrate LASSO and virtual-screening analysis.
   • catalyst-only-modeling/ for the catalyst-only model input and exported result plot.
4. DFT reaction-mechanism and kinetic-analysis materials are stored under DFT_reaction-mechanism/.

The folders are meant to be readable independently. Use the local README files in each workflow folder for run-order notes and file-specific details.

Descriptor Generation

descriptor_generation/ contains the raw descriptor provenance and compact tables used to build model-ready feature matrices.

Important files and folders:

• descriptor_generation/README: detailed guide for the descriptor folder.
• descriptor_generation/comb_input_template.xlsx: combined reaction template with catalyst, substrate, ddG, and descriptor columns.
• descriptor_generation/fill_paramters.py: fills descriptor columns in a combined workbook by matching catalyst and substrate IDs.
• descriptor_generation/catalysts/: catalyst ID workbook, Gaussian logs, and all_cats_descriptors.xlsx.
• descriptor_generation/substrates/: substrate ID workbook, Gaussian logs, and all_substrates_descriptors.xlsx.
• descriptor_generation/oos_catalysts/: out-of-sample catalyst descriptors.
• descriptor_generation/virtual_screening/: virtual-screening descriptor table for the substrate/analyte set.
• descriptor_generation/sisso/: SISSO-style reduced-feature inputs, outputs, scripts, and descriptor-name translation table.

To rebuild a combined descriptor table, run from descriptor_generation/:

python fill_paramters.py comb_input_template.xlsx catalysts/all_cats_descriptors.xlsx substrates/all_substrates_descriptors.xlsx

The script overwrites the input workbook, so work on a copy when preserving the template matters.

Catalyst-Substrate Modeling

catalyst-substrate-modeling/ contains the main notebook workflow:

• lasso_regression.ipynb: sequential LASSO modeling, validation, y-randomization, virtual-screening, and PCA analysis notebook.
• linear_features/: base linear feature train/test CSVs.
• linear_plus_steric_interaction_features/: interaction-feature train/test CSVs and saved result figures/tables.
• sisso_features/: SISSO feature train/test CSVs.
• out-of-sample-catalysts/: out-of-sample catalyst panel.
• virtual-screening/: default virtual-screening input set and saved outputs.

The notebook is sequential; run cells from top to bottom. The current detailed run configuration and expected outputs are documented in catalyst-substrate-modeling/README.md.

Catalyst-Only Modeling

catalyst-only-modeling/ contains the local inputs/results for the catalyst-only model:

• training_plus_val_set.xlsx: feature table with the summary_all_cats sheet.
• model_0_parity_publication.png: exported parity plot.

The modeling code itself is not vendored here. It uses the public SigmanGroup python-modeling workflow and the README documents how to use the workbook there.

DFT Reaction Mechanism And Kinetics

DFT_reaction-mechanism/ contains DFT structures, reaction-state summaries, transition-state conformer-search materials, and kinetic-analysis files.

Top-level DFT contents:

• PES_states_and_energies.xlsx: potential-energy-surface state and energy summary for the N-Me and 3,3'-Ph catalyst cases.
• N-Me_catalyst/: optimized structures, TS conformer-search results, selected conformer images, MD-run input, and a complete example workflow directory.
• 3_3prime_Ph_catalyst/: optimized structures, TS conformer summaries, selected images, and example Chemshell inputs.
• conformer_search_scripts/: reusable CREST/CENSO/Chemshell helper scripts. See DFT_reaction-mechanism/conformer_search_scripts/README.md.
• kinetic_simulations/: kinetic-analysis notebook, concentration-time workbook, fitted plots, and SI text. See DFT_reaction-mechanism/kinetic_simulations/README.md.

Environment

Create the local environment with:

conda env create -f betaine_env.yml
conda activate betaine_env

The environment includes Python 3.11, Jupyter kernel support, pandas, NumPy, scikit-learn, SciPy, matplotlib, seaborn, openpyxl, RDKit, DBSTEP, morfeus-ml, SHAP, Plotly, and notebook-rendering support packages.

External tools are still required for some workflows:

• Gaussian/GetProperties for regenerating descriptor tables from .log files.
• The SigmanGroup python-modeling repository for catalyst-only model execution.
• Cluster-specific CREST/CENSO/Chemshell/Turbomole tools for DFT conformer searches and optimizations.

License

Software source code and scripts in this repository are released under the MIT License. See LICENSE.

Data, figures, notebooks, molecular structures, calculation outputs, descriptor tables, model outputs, and documentation are released under the Creative Commons Attribution 4.0 International License (CC BY 4.0), unless otherwise noted. See DATA_LICENSE.md.

Third-party tools and external repositories referenced here remain subject to their own licenses.

Reproducibility Notes

• Read the folder-level README before rerunning notebooks or scripts.
• Several notebooks are sequential and reuse variables from earlier cells.
• Rerunning notebooks or helper scripts can overwrite plots, workbooks, and result summaries in the active output folder.
• This repository documents the current on-disk handoff state; it does not imply that all scientific computations were freshly rerun end-to-end.