PISA-T

Prediction of Interference with Specific Assay - Technologies is a project aimed at developing and evaluating machine learning models for predicting assay interference based on statistically derived labels from ultra-large bioactivity data matrices. This repository contains scripts for data preprocessing, model training, validation, and testing.

Installation

1. Clone the repository:

   git clone https://github.com/Bayer-Group/PISA-T.git

2. Install dependencies using Conda:

   conda env create -f environment.yml

Usage

1. Data Preparation

• Place your raw data files in the data/raw/ directory. | Note: data already available in the directory are randomly generated and serve as example to run the pipeline.
• Run the data preprocessing scripts in the preprocessing/ directory to clean and preprocess the data.

2. Model Training

• Use scripts in the dense_network/ and random_forest/ directories to train different models:
• Modify hyperparameters and configurations as needed.

3. Model Validation

• Validate the models using validation scripts provided in the respective directories.
• Tune hyperparameters for optimal performance.

4. Testing

• Test the trained models on test data using testing scripts.
• Evaluate model performance and generate results.

Contributors

• Vincenzo Palmacci (@vincenzo-palmacci)

License

This project is licensed under the BSD 3-Clause License - see the LICENSE file for details.