ML_Models

This repository collects the notebook workflows, local code dependencies, and result artifacts used for graph-based DL-ML modeling work for PK parameters prediction and benchmarking.

Shared code

• AttentiveFP/: local Python package required by the Jupyter notebooks
• config.py: central training and path configuration
• utils.py: shared helper functions for training/evaluation flow
• requirements.txt: Python dependency list
• README.md: notes on what to keep with the notebook workflow

Main notebook cases

• DL_train_lgCL.ipynb: the CL deep-learning training notebook that imports AttentiveFP
• embML_CL_SVR.ipynb: embedding + SVR workflow for lgCL evaluation

Data files in this repo

• data_prep/df_feature_5620.csv: main input table used by training notebook
• data_prep/lgCL_Embeddings_RDKIT_S1.csv
• data_prep/lgFu_Embeddings_RDKIT_S1.csv
• data_prep/lgVD_Embeddings_RDKIT_S1.csv

Local cache file (not required for clean sharing/reproduction):

• data_prep/df_feature_5620.pickle.pickle

Notes

• Keep generated checkpoints, checkpoint caches, and other large intermediates out of Git unless they are part of the deliverable.
• Use .gitignore to avoid accidental commits of notebook caches, Python bytecode, and temporary artifacts.