An information theoretic approach for analyzing temporal patterns of gene expression.

Area of Science:

• Bioinformatics
• Computational Biology
• Systems Biology

Background:

• Microarray technology enables simultaneous measurement of thousands of mRNA expression levels.
• Gene expression data is rich but requires advanced mining for biological insights.
• Information theoretic methods can quantify similarities and dissimilarities in data distributions.

Purpose of the Study:

• To investigate information theoretic data mining approaches for discovering temporal gene expression patterns.
• To apply these methods to array-derived gene expression data.

Main Methods:

• Utilized Kullback-Leibler (KL) divergence, an information-theoretic measure of distribution dissimilarity.
• Employed an unsupervised self-organizing map (SOM) algorithm in conjunction with KL divergence.
• Analyzed two published, array-derived gene expression datasets.

Main Results:

• KL divergence-based clustering identified superior temporal patterns compared to hierarchical clustering with Pearson correlation.
• The discovered patterns demonstrated biological significance upon examination.
• The KL clustering method proved more effective in uncovering meaningful temporal trends in gene expression.

Conclusions:

• Information theoretic methods, specifically KL divergence with SOM, are powerful tools for analyzing temporal gene expression patterns.
• This approach offers advantages over traditional correlation-based clustering methods for microarray data.
• The findings highlight the utility of advanced computational techniques in extracting biological knowledge from high-throughput experiments.