Recursive fuzzy granulation for gene subsets extraction and cancer classification.

Area of Science:

• Bioinformatics
• Computational Biology
• Machine Learning in Genomics

Background:

• Microarray gene expression datasets are often sparse and imbalanced, posing challenges for accurate cancer classification.
• Identifying highly informative gene subsets is crucial for developing robust diagnostic tools.

Purpose of the Study:

• To introduce a novel hybrid algorithm, Fuzzy Granular Support Vector Machine—Recursive Feature Elimination (FGSVM-RFE), for selecting optimal gene subsets from complex gene expression data.
• To enhance cancer classification and diagnosis by improving the accuracy and efficiency of gene selection.

Main Methods:

• The FGSVM-RFE algorithm integrates statistical learning, fuzzy clustering, and granular computing.
• It iteratively eliminates irrelevant, redundant, or noisy genes within different data granules.
• The method selects biologically meaningful gene subsets with distinct functions.

Main Results:

• Empirical studies on three public datasets confirmed that FGSVM-RFE surpasses current state-of-the-art methods.
• The algorithm successfully extracted multiple gene subsets, each enabling 100% accurate classifier modeling.
• For prostate cancer dataset, independent testing accuracy improved from 86% with 16 genes to 100% with only eight genes.

Conclusions:

• FGSVM-RFE offers a powerful and accurate approach for gene selection in cancer research.
• The identified gene subsets are biologically validated, suggesting clinical relevance for cancer diagnosis and classification.
• This method significantly advances the field of genomic data analysis for improved disease detection.