[Cluster ensemble algorithm based on dual neural gas applied to cancer gene expression profiles]

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Microarray technology is crucial for biological and medical research, offering insights into cancer diagnosis and treatment.
• Accurate classification of cancer samples and identification of diverse cancer types are critical challenges in oncology.

Purpose of the Study:

• To propose a novel cluster ensemble framework, Dual Neural Gas Cluster Ensemble (DNGCE), for analyzing noisy cancer gene expression profiles.
• To enhance the discovery of underlying data structures for improved cancer subtyping and sample classification.

Main Methods:

• The DNGCE framework utilizes the neural gas algorithm for clustering on both sample and attribute dimensions.
• It incorporates the normalized cut algorithm to partition a consensus matrix derived from multiple clustering outcomes.
• The approach is designed to handle the inherent noise in gene expression data.

Main Results:

• Experiments on cancer gene expression profiles demonstrate the effectiveness of the DNGCE framework.
• The proposed approach achieves superior performance compared to single clustering algorithms.
• DNGCE outperforms most existing methods for clustering gene expression profiles.

Conclusions:

• The DNGCE framework provides accurate results for cancer sample classification and subtype discovery.
• This novel approach offers a significant advancement in analyzing complex genomic data for cancer research.
• The findings suggest DNGCE's potential to aid in developing more precise cancer diagnostics and treatments.