COSINE: COndition-SpecIfic sub-NEtwork identification using a global optimization method.

Area of Science:

• Bioinformatics
• Computational Biology
• Systems Biology

Background:

• Identifying condition-specific gene sub-networks is crucial for understanding disease mechanisms and discovering biomarkers.
• Existing methods often overlook valuable information by not simultaneously analyzing individual gene expression and gene pair correlations.

Purpose of the Study:

• To introduce COSINE (COndition SpecIfic sub-NEtwork), a novel computational method for extracting condition-specific sub-networks.
• To address the limitations of current approaches by integrating differential gene expression and differential gene pair correlation.

Main Methods:

• COSINE utilizes a scoring function that quantifies condition-specific changes in both individual genes (nodes) and their co-expression relationships (edges).
• A genetic algorithm is employed to efficiently search for the optimal sub-network that maximizes this scoring function.
• The method was validated on simulated datasets and real-world data, including prostate cancer, obesity, and HapMap samples.

Main Results:

• COSINE demonstrated superior performance compared to existing methods in identifying significant sub-networks.
• The identified sub-networks were of appropriate size and possessed meaningful biological relevance.
• The method successfully captured condition-specific alterations in gene expression and co-expression patterns.

Conclusions:

• COSINE provides a powerful and effective approach for the identification of condition-specific gene sub-networks.
• By integrating node and edge differential information, COSINE enhances the biological interpretability and significance of discovered sub-networks.
• The COSINE R package is available on CRAN, facilitating its application in biological research.