Combining hierarchical clustering and self-organizing maps for exploratory analysis of gene expression patterns

Area of Science:

• Computational biology
• Bioinformatics
• Data science

Background:

• Self-organizing maps (SOM) provide linear run times and handle noisy data effectively, presenting an alternative to classical clustering.
• A limitation of SOM is its dependence on the relative sizes of the clusters being analyzed.
• Massive datasets, such as those from DNA microarray expression patterns, present unique analytical challenges.

Purpose of the Study:

• To investigate the effectiveness of combining Self-organizing maps (SOM) with hierarchical clustering.
• To determine if this combined approach enhances exploratory data analysis for large, complex datasets.
• To address the limitations of SOM in handling varying cluster sizes.

Main Methods:

• The study employed Self-organizing maps (SOM) as a primary clustering technique.
• Hierarchical clustering methods were integrated with SOM to refine the analysis.
• The combined approach was applied to analyze DNA microarray expression patterns.

Main Results:

• The combination of SOM and hierarchical clustering proved to be a powerful tool for exploratory analysis.
• This integrated method demonstrated superior performance in handling massive datasets compared to SOM alone.
• The approach effectively addressed the cluster size dependency issue inherent in standalone SOM.

Conclusions:

• The integration of Self-organizing maps (SOM) with hierarchical clustering offers a robust solution for analyzing large-scale biological data.
• This hybrid methodology enhances the accuracy and reliability of exploratory data analysis, particularly for noisy datasets like DNA microarrays.
• The combined approach provides a valuable tool for uncovering patterns in complex biological expression data.