About

This is project that illustrates end-to-end development of a Thrombin inhibiton predictive model. The data was acquired from the Molecule Activity Cliff Estimation (MoleculeACE) CHEMBL204_Ki benchmark dataset available here.

Make sure to set the correct MLflow Tracking URI: This can be done in the main.py by adding mlflow.set_tracking_uri(PATH__TO__FOLDER), where PATH__TO__FOLDER can be the mlruns/ folder under the root directory of the project. You can check the .mlflow configuration directory or your environment variables: echo $MLFLOW_TRACKING_URI, and echo $MLFLOW_ARTIFACT_URI.

Covered topics include:

1. Data preprocessing (featurization, cleaning, splitting),
2. Model training (via hyperparameter tuning) and validation,
3. ML experiment tracking, model regitry and serving using MLflow

Installation

1. Create a conda environment conda create -n thrombin python=3.10
2. Install required packages by running pip install -r requirements.txt

Model Training and validation

Run the main app for data acquisition/preprocessing, and model training/evaluation/registration

1. python main.py