A modified binary particle swarm optimization with a machine learning algorithm and molecular docking for QSAR...

Area of Science:

• Computational Chemistry
• Medicinal Chemistry
• Pharmacology

Background:

• Acetylcholinesterase (AChE) and butyrylcholinesterase (BuChE) inhibitors are crucial for Alzheimer's disease treatment.
• Quantitative Structure-Activity Relationship (QSAR) models are essential for predicting drug efficacy.
• Existing QSAR methods require optimization for descriptor selection and model building.

Purpose of the Study:

• To develop an advanced QSAR modeling approach for predicting AChE and BuChE inhibitor activity.
• To integrate modified binary particle swarm optimization (PSO) with machine learning for enhanced descriptor selection.
• To improve the exploration capability for identifying optimal descriptor subsets for QSAR models.

Main Methods:

• A modified binary particle swarm optimization (PSO) algorithm was employed, incorporating a transfer function for binary conversion.
• Catfish effect and chaotic maps were utilized to enhance the exploration ability of PSO.
• Various machine learning algorithms (K-nearest neighbor, multiple linear regression, support vector machine, regression tree) were combined with PSO for QSAR model construction and validation.

Main Results:

• The integrated approach successfully built QSAR models to predict the activity of AChE and BuChE inhibitors.
• Different combinations of transfer functions and machine learning algorithms yielded various model performances.
• Internal and external validation confirmed the reliability of the constructed QSAR models.

Conclusions:

• The proposed QSAR modeling strategy effectively predicts enzyme inhibitor activity for Alzheimer's disease drug discovery.
• The integration of modified binary PSO with machine learning offers a robust framework for developing predictive pharmacological models.
• The study highlights the potential of optimized descriptor selection for enhancing QSAR model accuracy.