Citation

This repo contains the code and datasets for the publication

G.A. Landrum, M. Beckers, J. Lanini, N. Schneider, N. Stiefl, S. Riniker 
"SIMPD: an Algorithm for Generating Simulated Time Splits for Validating Machine Learning Approaches"
J. Cheminform. 15, 119 (2023). https://doi.org/10.1186/s13321-023-00787-9

Data and results

Directories:

• datasets: contains the Ki datasets exported from ChEMBL32 as well as the results of running SIMPD on the assay datasets. There's also a data catalog description for the intake package in here

Jupyter Notebooks

• 01_Get_ChEMBL32_Bioactivity_Data_assays.ipynb : export the raw ChEMBL Ki datasets from ChEMBL
• 02_ChEMBL_ML_experiments.ipynb : analysis of the results of the ML experiments on the ChEMBL Datasets

Python files

• ga_lib_3.py : Python library with utility functionality for implementing SIMPD and doing the analysis. This is used in most of the scripts and juypter notebooks

Other

• pymoo.yml : conda environment specification for the environment that was used during this project
• run_chemblassay_ml_nogap.py : script to run the ML experiments on the base assay data
• run_chemblassay_ga_nogap.py : script to run SIMPD on the base assay data
• run_chemblassay_ml_moo_nogap.py : script to run the ML experiments on the SIMPD results
• create_simpd_output_chemblKi.py : script to create CSV files with the final SIMPD datasets. This is what ends up in the datasets/chembl_SIMPD directory