KUALA: Kinase drUgs mAchine Learning frAmework

Multi-target priority scores are contained in the results folder. Each row contains:

• Kinase Uniprot ID
• ChEMBL ID of the ligand predicted as active
• MTPS score for kinase-ligand pair
• Status of the repurposing choice

Machine learning models

Trained models to predict activity for our set of kinases can be downloaded from Zenodo https://doi.org/10.5281/zenodo.6554043

Tutorial

KUALA models have been trained by using molecular descriptors computed with PaDEL-Descriptor software.

The following steps are useful to correctly predict activity for our set of kinases:

1. Download KUALA models from Zenodo: https://doi.org/10.5281/zenodo.6867485 (PaDEL-Descriptors is also available as an extension of KNIME)
2. Compute all available molecular descritors (a comprehensive list of mandatory descriptors is reported in mandatory-list.txt)
3. Set path variables properly within kuala-demo.R script and execute it
4. Collect your results, default contained in kuala_predicted_ligands_activity.txt file in your current working directory

Results

A comprehensive overview of repurposable drugs for each kinase is reported on figure below and is freely accessible on shinyapps.io.