Cheminformatics Analysis and Modeling with MacrolactoneDB.

Area of Science:

• Medicinal Chemistry
• Cheminformatics
• Drug Discovery

Background:

• Macrolactones are macrocyclic lactones with a core ring of at least twelve atoms.
• They encompass diverse natural products, including macrolides, known for potent bioactivities like antibiotics and favorable drug-like properties.
• This structural class holds significant therapeutic potential but remains underexplored.

Purpose of the Study:

• To develop MacrolactoneDB, a database integrating existing macrolactone structures and bioactivity data.
• To introduce novel molecular descriptors for enhanced characterization of macrolide structures.
• To analyze the chemical distribution and therapeutic potential of macrolactones using cheminformatics approaches.

Main Methods:

• Integrated nearly 14,000 macrolactones and their bioactivity data from public databases into MacrolactoneDB.
• Developed new molecular descriptors to characterize macrolide structures.
• Employed regression machine learning models with seven descriptor sets and eight algorithms to predict biological endpoints for targets including Plasmodium falciparum, Hepatitis C virus, and T-cells.

Main Results:

• Analyzed the chemical distribution of MacrolactoneDB based on key molecular properties.
• Demonstrated the value of compiled data for understanding macrolactone bioactivity against specific disease targets.
• Identified that merging molecular descriptors significantly enhances predictive power, particularly with Random Forest models, often improved by consensus or hybrid approaches.

Conclusions:

• MacrolactoneDB provides valuable cheminformatics insights into this privileged structural class.
• The integration of diverse data and advanced descriptors facilitates the prediction of biological activity.
• Macrolactones represent a promising, yet underexplored, area for drug discovery with high therapeutic potential.