Gene selection for cancer identification: a decision tree model empowered by particle swarm optimization algorithm.

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Selecting informative genes from large microarray datasets is crucial for cancer research.
• Computational intelligence methods are increasingly used for analyzing gene expression data.

Purpose of the Study:

• To develop a novel gene selection method for cancer identification using particle swarm optimization and decision trees.
• To compare the proposed method's performance against established classification techniques.

Main Methods:

• Utilized particle swarm optimization (PSO) integrated with a decision tree classifier.
• Conducted comparative experiments on 11 gene expression cancer datasets.
• Evaluated against benchmark methods including Support Vector Machine (SVM), Self-Organizing Map (SOM), and various decision tree algorithms.

Main Results:

• The proposed PSO-decision tree method demonstrated superior performance across all tested cancer datasets.
• The method achieved comparable results to SVM on specific datasets.
• Identified key genes, including housekeeping and tissue-specific genes, with high cancer classification power.

Conclusions:

• The novel PSO-decision tree approach offers an effective solution for gene selection in cancer classification.
• Identified genes provide significant discriminatory power for cancer subtypes.
• This method enhances the ability to identify informative genes from complex genomic data.