An Integrated Feature Selection Algorithm for Cancer Classification using Gene Expression Data.

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Cancer is a global health threat driven by somatic genomic mutations.
• Early cancer diagnosis is crucial for effective treatment.
• DNA microarray technology generates high-dimensional gene expression data, posing challenges for analysis.

Purpose of the Study:

• To develop an efficient and robust feature selection method for cancer diagnosis.
• To identify a minimal set of informative genes for improved classification performance.
• To enhance early cancer detection using gene expression data.

Main Methods:

• A hybrid feature selection method integrating Correlation-Based Feature Selection (CFS) and Multi-Objective Evolutionary Algorithm (MOEA) was developed.
• The hybrid model utilized a Radial Basis Function Neural Network (RBFNN) classifier.
• Performance was evaluated on 11 benchmark gene expression datasets using 10-fold cross-validation.

Main Results:

• The proposed CFS-MOEA algorithm demonstrated superior performance compared to seven conventional feature selection methods.
• The hybrid approach achieved high classification accuracy ratios.
• The method identified a minimal subset of predictive genes.

Conclusions:

• The CFS-MOEA hybrid algorithm is effective for feature selection in cancer gene expression data analysis.
• The method achieved up to 100% classification accuracy on 6 out of 11 datasets.
• This approach offers a promising tool for early cancer diagnosis and biomarker discovery.