Combining machine learning and structure-based approaches to develop oncogene PIM kinase inhibitors.

Area of Science:

• Medicinal Chemistry
• Computational Biology
• Drug Discovery

Background:

• PIM kinases are implicated in various cancers by promoting cell survival and proliferation.
• There is a demand for potent PIM kinase inhibitors with favorable pharmacologic profiles for cancer therapy.

Purpose of the Study:

• To generate novel and effective chemical therapeutics for PIM-1 kinase using integrated approaches.
• To identify potent PIM-1 inhibitors with potential for cancer treatment.

Main Methods:

• Employed machine learning (SVM, Random Forest, k-NN, XGBoost) and structure-based techniques.
• Utilized Boruta method for descriptor selection and an ensemble approach for model development.
• Validated potential inhibitors using molecular docking and molecular dynamics simulations.

Main Results:

• Support Vector Machine (SVM), Random Forest, and XGBoost models demonstrated superior performance over k-NN.
• Identified four potential PIM-1 modulating molecules: CHEMBL303779, CHEMBL690270, MHC07198, and CHEMBL748285.
• Molecular dynamics simulations confirmed the stability of protein-ligand interactions for the identified molecules.

Conclusions:

• The developed machine learning models are robust and effective for PIM kinase inhibitor discovery.
• The identified molecules hold potential for the development of novel PIM kinase inhibitors against human cancers.