Extraction of correlated gene clusters by multiple graph comparison

Area of Science:

• Bioinformatics
• Computational Biology
• Systems Biology

Background:

• Genes exhibit relationships based on genomic position, functional pathways, and expression profiles.
• Correlated gene clusters across multiple biological features suggest stronger functional links.
• Existing methods may not fully capture these multi-graph correlations.

Purpose of the Study:

• To develop a new heuristic algorithm for extracting correlated gene clusters.
• To identify genes with shared relevance across diverse biological datasets.
• To assess the method's utility in analyzing large biological interaction datasets.

Main Methods:

• Encoding gene relationships as graph structures (genome, pathway, expression profiles).
• Defining inter-graph links based on biological relevance.
• Extracting correlated gene clusters as isomorphic subgraphs across multiple graphs.

Main Results:

• Identified correlated gene clusters in E.coli based on genomic position, metabolic pathways, and 3D structure.
• Applied the method to analyze protein-protein interaction and gene coexpression data in S.cerevisiae.
• Demonstrated the potential for screening datasets to identify reliable biological relationships.

Conclusions:

• The heuristic algorithm effectively extracts biologically relevant correlated gene clusters.
• The method aids in integrating and interpreting multi-modal biological data.
• This approach can help in identifying true positive interactions and reducing false positives in large-scale datasets.