A novel feature selection method for microarray data classification based on hidden Markov model.

Area of Science:

• Bioinformatics
• Computational Biology
• Machine Learning

Background:

• Feature selection is crucial for effective cancer classification from high-dimensional microarray data.
• Integrating diverse feature selection methods can enhance classification performance.
• Existing methods may not optimally combine multiple criteria for robust gene selection.

Purpose of the Study:

• To develop a novel approach for integrating multiple feature selection criteria using a hidden Markov model (HMM).
• To evaluate the performance of the proposed HMM-based gene selection strategy in cancer classification.
• To compare the HMM approach against other gene selection and rank aggregation methods.

Main Methods:

• A hidden Markov model (HMM) topology was designed to integrate five feature selection ranking methods: Bhattacharyya distance, entropy, receiver operating characteristic curve, t-test, and Wilcoxon.
• A strategy for constructing, learning, and inferring the HMM for gene selection was presented.
• The proposed method was evaluated on three publicly available microarray datasets for diffuse large B-cell lymphoma, leukemia, and prostate cancer.

Main Results:

• The HMM-based gene selection demonstrated superior performance in cancer classification compared to Markov chain rank aggregation.
• The proposed approach outperformed individual feature selection criteria when applied to general classifiers.
• The method showed higher classification accuracy on diffuse large B-cell lymphoma, leukemia, and prostate cancer datasets.

Conclusions:

• The proposed HMM-based approach offers a powerful procedure for combining different feature selection methods.
• This integrated strategy leads to more robust gene selection and improved classification performance in real-world applications.
• The HMM framework provides an effective way to leverage multiple criteria for enhanced biological data analysis.