Novel implementation of conditional co-regulation by graph theory to derive co-expressed genes from microarray data

Area of Science:

• Bioinformatics
• Systems Biology
• Genomics

Background:

• Existing transcriptional databases rely on correlation methods (Pearson, Spearman) to identify gene co-regulation.
• These methods often exclude genes not expressed across all conditions, leading to incomplete co-regulation analysis.
• This limitation hinders the study of gene relationships within specific pathways or conditions.

Purpose of the Study:

• To develop an alternative method for analyzing large transcriptional datasets using graph theory.
• To overcome the limitations of existing databases by including all genes, regardless of expression levels or conditions.
• To identify novel, biologically relevant gene relationships and co-regulations.

Main Methods:

• Implemented a graph theory-based algorithm to analyze time-series microarray data (AtGenExpress).
• Converted discretized gene expression signals into strings to represent gene behavior across conditions.
• Calculated gene relationships using a similarity index based on string matching, generating a 'score' for each gene pair.

Main Results:

• The method successfully identified genes involved in similar functions and pathways, as demonstrated in the carbohydrate metabolism test case.
• It recognized correlated gene pairs from existing databases (CSB.DB) and uncovered additional potentially important relationships.
• All genes, irrespective of expression values or condition-specific expression, were included, highlighting previously missed connections.

Conclusions:

• Conditional co-regulation analysis using graph theory effectively identifies novel gene relationships in large expression datasets.
• The method's ability to include all genes, even those with low or conditional expression, provides a more comprehensive understanding of gene interactions.
• This approach addresses limitations of traditional clustering methods and offers a powerful tool for biological discovery, exemplified by the ASIDB database.