Visualization and evaluation of clusters for exploratory analysis of gene expression data

Area of Science:

• Bioinformatics
• Computational Biology
• Data Science

Background:

• Clustering algorithms are vital for analyzing large-scale gene expression profiles.
• Effective visualization and objective cluster evaluation are often neglected in current methods.
• There is a need for a standardized framework to develop robust clustering algorithms.

Purpose of the Study:

• To propose a theoretical framework and formalizations for consistent development of clustering algorithms.
• To introduce a novel clustering algorithm within this framework.
• To enhance cluster visualization and objective evaluation in gene expression data analysis.

Main Methods:

• Development of a theoretical framework for clustering algorithm design.
• Introduction of a new clustering algorithm based on the framework.
• Implementation of comprehensive data visualization and objective cluster quality measures.
• Rigorous evaluation against established algorithms like K-means and Self-Organizing Maps.

Main Results:

• The new algorithm, developed within the proposed framework, demonstrates improved cluster consistency and quality.
• Comparative analysis on four gene expression datasets shows promising results against popular methods.
• The framework facilitates uniform application of principles for optimization, visualization, and evaluation.

Conclusions:

• The proposed framework provides a sound theoretical basis for developing advanced clustering algorithms.
• The novel algorithm offers superior performance in terms of cluster quality and visualization for gene expression data.
• This work addresses critical needs in objective evaluation and visualization for clustering in bioinformatics.