Iterative ensemble feature selection for multiclass classification of imbalanced microarray data.

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Microarray technology enables monitoring of thousands of gene expression levels across diverse tumor tissues.
• Accurate classification of tumor types using gene expression data is crucial for clinical studies.
• Multiclass microarray data often exhibit imbalanced class distributions, hindering the performance of One-Versus-All (OVA) classification.

Purpose of the Study:

• To propose a novel iterative ensemble feature selection (IEFS) framework.
• To address the challenge of imbalanced class distribution in multiclass microarray data classification.
• To enhance the performance of the One-Versus-All (OVA) classification schema.

Main Methods:

• Developed an iterative ensemble feature selection (IEFS) framework.
• Integrated filter feature selection and balanced sampling iteratively within the OVA schema.
• Evaluated the IEFS framework against state-of-the-art methods on six benchmark multiclass microarray datasets.

Main Results:

• The IEFS framework demonstrated superior or comparable performance compared to existing methods.
• Performance was measured by classification accuracy and area under the receiver operating characteristic curve.
• The IEFS framework's effectiveness increased with a higher number of classes in the data.

Conclusions:

• The combination of balanced sampling and feature selection significantly improves multiclass classification of imbalanced microarray data.
• The proposed IEFS framework offers a robust solution for analyzing imbalanced biological datasets.
• The IEFS framework is adaptable for other biological data analysis tasks with similar challenges.