Automated drug design for druggable target identification using integrated stacked autoencoder and hierarchically...

Area of Science:

• Pharmaceutical Informatics
• Computational Biology
• Machine Learning in Drug Discovery

Background:

• Drug classification and target identification are critical but challenging in drug discovery.
• Existing methods like SVMs, XGBoost, and deep learning models face limitations in efficiency, scalability, interpretability, and generalization.
• There is a need for advanced computational frameworks to handle complex pharmaceutical data.

Purpose of the Study:

• To introduce a novel computational framework, optSAE+HSAPSO, for efficient and accurate drug classification and target identification.
• To address the limitations of existing methods in terms of accuracy, computational complexity, and scalability.
• To provide a robust and adaptable solution for real-world drug discovery applications.

Main Methods:

• Integration of a stacked autoencoder (SAE) for robust feature extraction.
• Utilization of a hierarchically self-adaptive particle swarm optimization (HSAPSO) algorithm for adaptive parameter optimization.
• Experimental evaluation on DrugBank and Swiss-Prot datasets.

Main Results:

• The optSAE+HSAPSO framework achieved a high accuracy of 95.52%.
• Demonstrated significantly reduced computational complexity (0.010 s/sample) and exceptional stability (±0.003).
• Outperformed state-of-the-art methods in accuracy, convergence speed, and resilience to variability.

Conclusions:

• The optSAE+HSAPSO framework offers a scalable, adaptable, and efficient solution for drug classification and target identification.
• The framework shows robustness and generalization capabilities, maintaining consistent performance on validation and unseen datasets.
• This work advances pharmaceutical informatics and accelerates drug development, with potential applications in disease diagnostics and genetic data classification.