Machine Learning Application for Medicinal Chemistry: Colchicine Case, New Structures, and Anticancer Activity...

Area of Science:

• Computational Chemistry
• Medicinal Chemistry
• Machine Learning in Drug Discovery

Background:

• Traditional quantitative structure-activity relationship (QSAR) methods are enhanced by machine learning (ML) for drug discovery.
• Colchicine derivatives are explored for their anticancer potential.
• Predictive modeling is crucial for identifying effective and safe drug candidates.

Purpose of the Study:

• To assess the anticancer potential of colchicine-based compounds across five cell lines.
• To develop ML models for predicting anticancer activity (IC50 values) and generating novel compounds.
• To compute resistance index (RI) and selectivity index (SI) for drug resistance and safety evaluation.

Main Methods:

• Utilized various ML algorithms including random forest, decision tree, support vector machines, k-nearest neighbors, and multiple linear regression.
• Assessed inhibitory capabilities against five cancer cell lines.
• Developed a novel ML system to recommend new chemical structures based on known anticancer activity.

Main Results:

• Established a library of novel colchicine-based compounds with predicted IC50 values.
• Developed a validated predictive ML model capable of estimating IC50 values from molecular structures.
• Identified promising colchicine derivatives with potential anticancer activity.

Conclusions:

• Machine learning effectively predicts anticancer activity and aids in the design of novel colchicine-based compounds.
• The developed predictive model provides a reliable tool for estimating drug efficacy based on molecular structure.
• This study facilitates the discovery of new anticancer agents with improved therapeutic profiles.