MolToxPred

MolToxPred is a machine learning based tool to predict toxicity scores of small molecules

Prerequisites:

• Python 3.7+

• Java JRE 6+

• requirements.txt file contains all the necessary python packages required.

Usage:

To use the trained models for predictions:

1. Download and unpack the zip file/ Clone the GitHub library

2. Create an environment with dependencies using requirements.txt file

3. Prepare your input file, the molecules should be in SMILES format. For single molecule SMILES can be entered directly, for multiple molecules prepare a .csv file

4. Run the MolToxPred by python main_moltox.py

5. The output file is generated as your 'custom input_results.csv'

The main_moltox.py file will generate descriptors using RDKit and molecular fingerprints using Padelpy for a molecule. fingerprints_xml.zip will be parsed to generate the fingerprints,feature_list.pkl will perform the feature selection as described in the manucript and output will be individual fingerprint file with selected features in Padel folder. Toxicity prediction of the molecule will happen using the trained model stacked_model.jobliband results.csv output will be created with your custom file name having toxicity score.

Datasets:

An example test set that can be used for prediction (in .csv format) is provided in sample_SMILES.