A new measure for gene expression biclustering based on non-parametric correlation.

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• DNA microarray analysis uses biclustering to find co-expressed gene subsets, revealing biological processes.
• Existing methods often miss complex patterns like shifting, scaling, or gene inversions crucial for understanding cancer.

Purpose of the Study:

• To introduce a novel biclustering measure, Spearman's biclustering measure (SBM), for enhanced DNA microarray data analysis.
• To develop a method capable of detecting complex coherence patterns missed by traditional approaches.

Main Methods:

• Spearman's biclustering measure (SBM) estimates bicluster quality using non-linear correlation across genes and conditions.
• An estimation of distribution algorithm, employing SBM as its fitness function, was used for bicluster searching.
• The approach was validated using artificial and real microarrays, quality indexes, reference patterns, and statistical tests.

Main Results:

• SBM successfully identifies complex bicluster patterns including shifting, scaling, and inversion.
• The method demonstrated effective performance on real microarray data, outperforming other algorithms like Bimax, CC, OPSM, Plaid, and xMotifs.
• SBM allows for selective marginalization of genes and conditions based on statistical significance.

Conclusions:

• SBM offers significant advantages over existing methods by detecting a wider range of coherence patterns in gene expression data.
• The ability to capture complex relationships enhances the discovery of subtle biological processes, particularly in cancer research.
• SBM provides a more nuanced and statistically robust approach to biclustering DNA microarray data.