DQB: A novel dynamic quantitive classification model using artificial bee colony algorithm with application on gene...

Area of Science:

• Medical informatics
• Bioinformatics
• Machine learning

Background:

• Developing comprehensible rule-based classification models is crucial in medicine for understanding prediction drivers.
• Existing methods often lack interpretability, hindering biological insights from complex datasets.

Purpose of the Study:

• To propose a novel dynamic quantitative rule-based classification model (DQB) for enhanced understandability and accuracy.
• To apply DQB for classifying microarray gene expression profiles, a first in this domain.
• To introduce a new dynamic local search (DLS) strategy to improve the Artificial Bee Colony (ABC) algorithm.

Main Methods:

• Integrating quantitative association rule mining with the Artificial Bee Colony (ABC) algorithm.
• Developing a new dynamic local search (DLS) strategy to optimize the ABC algorithm.
• Evaluating the DQB model on six gene expression datasets for cancer classification.

Main Results:

• DQB achieved a considerable increase in classification accuracy compared to existing methods.
• The model provides an interpretable rule-based system valuable for biologists.
• Quantitative rules achieved near 100% accuracy with minimal genes, suggesting high-quality knowledge extraction.
• Several potentially novel genes were identified through this analysis.

Conclusions:

• The DQB model offers an accurate and interpretable approach for gene expression profile classification.
• The method demonstrates the potential for discovering meaningful biological knowledge and novel biomarkers.
• DQB shows promise for real-world applications beyond gene expression analysis with modifications.