DOMINO: a network-based active module identification algorithm with reduced rate of false calls.

Area of Science:

• Bioinformatics
• Computational Biology
• Systems Biology

Background:

• Active module identification (AMI) algorithms are crucial for analyzing omics data, aiming to pinpoint biological processes driving specific conditions.
• Current AMI methods often identify modules lacking specificity, with enriched terms appearing in random data permutations.

Purpose of the Study:

• To systematically evaluate existing AMI methods.
• To develop a bias-correction method for evaluating identified modules.
• To introduce a novel, high-performing AMI algorithm.

Main Methods:

• Systematic evaluation of six popular AMI algorithms using gene expression and GWAS data.
• Development of a permutation-based method to empirically assess Gene Ontology (GO) term enrichment.
• Creation of five new AMI performance criteria.
• Development and testing of the DOMINO AMI algorithm.

Main Results:

• A significant bias was observed where GO terms enriched in modules from real data were also enriched in modules from permuted data.
• The novel permutation-based method effectively evaluates GO term specificity.
• The DOMINO algorithm demonstrated superior performance compared to six other AMI algorithms in extensive testing.

Conclusions:

• Existing AMI methods may produce non-specific results, potentially misrepresenting biological insights.
• The developed permutation-based evaluation and the DOMINO algorithm offer improved accuracy and specificity in active module identification.
• DOMINO provides a robust tool for analyzing omics data, enhancing the discovery of biologically relevant subnetworks.