Analysis and visualization of gene expression data using self-organizing maps.

Area of Science:

• Bioinformatics
• Computational Biology
• Systems Biology

Background:

• Gene expression data analysis is crucial for understanding biological systems.
• Identifying clusters and relationships within this data aids in hypothesis generation.
• Existing methods may have limitations in visualizing complex similarity structures.

Purpose of the Study:

• To analyze and visualize the cluster structure of gene expression data using the Self-Organizing Map (SOM) algorithm.
• To demonstrate SOM's utility as an exploratory data analysis tool for gene function discovery.
• To compare SOM's visualization of gene similarity against alternative methods.

Main Methods:

• Application of the Self-Organizing Map (SOM) algorithm to gene expression data from DNA microarrays.
• Utilizing the U-matrix method for visualizing cluster structures.
• Characterizing gene clusters based on expression profile properties.
• Comparative analysis with Multidimensional Scaling (MDS) and Hierarchical Clustering (HC).

Main Results:

• SOM generates a non-linear, two-dimensional map of gene expression data.
• The distribution of known functional gene classes on the SOM reveals relationships.
• The U-matrix effectively visualizes the cluster structure.
• SOM demonstrates a more trustworthy visualization of gene similarity compared to MDS and HC.

Conclusions:

• The Self-Organizing Map (SOM) is a powerful tool for exploratory analysis of gene expression data.
• SOM facilitates hypothesis generation regarding gene relationships and functions.
• SOM provides a superior and more reliable visualization of gene similarity in expression datasets.