A topology-preserving selection and clustering approach to multidimensional biological data

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Multidimensional genome-wide data, such as gene expression microarrays, offer extensive biological insights.
• The inherent relationships within these complex datasets are often overlooked.
• The spatial properties of data clouds in hyperspace can reveal underlying biological relationships.

Purpose of the Study:

• To introduce an analytical improvement for complex, large-scale microarray data analysis.
• To develop a method that preserves data topology for more accurate biological interpretation.
• To facilitate the identification of gene clusters with similar expression patterns for functional and regulatory relevance.

Main Methods:

• Development of a topology-preserving selection and clustering (TPSC) approach.
• Integration of self-organizing map (SOM) and singular value decomposition (SVD) for genome-wide selection.
• Application of an SOM-based two-phase gene clustering procedure for topology-preserving identification.

Main Results:

• The TPSC method demonstrated superior performance in extracting characteristic features from large, complex datasets.
• Successful application to human cell cycle, stress response, and host-pathogen interaction datasets.
• Identified gene clusters with highly similar expression patterns, aiding biological interpretation.

Conclusions:

• The topology-preserving selection and clustering approach enables in-depth, unbiased biological information mining.
• TPSC does not require a priori assumptions about data structure.
• The method expands the scope of omics applications and is available via a web server.