QDB (Query Driven Biclustering)

QDB performs Bayesian query-driven biclustering of microarray gene expression data to identify functionally related gene modules and the experimental conditions in which they co-express, using seed genes as priors.

Key Features:

• Query-Driven Seeding: Uses seed genes believed to share a common function to recruit additional genes with similar expression patterns across subsets of experimental conditions.
• Bayesian Framework: Integrates prior knowledge from seed genes into a probabilistic model to guide biclustering searches.
• Resolution Sweep Approach: Grows biclusters from small seed sets across a range of resolutions to identify functionally enriched and modular clusters, including overlapping modules.
• Missing Value Handling: Natively accommodates missing values in microarray data within its probabilistic framework.
• Performance Validation: Validated on yeast compendia and artificial data from recent biclustering benchmark studies, demonstrating advantages for focused module search.

Scientific Applications:

• Targeted module discovery: Identifies gene modules guided by prior biological knowledge represented by seed genes.
• Condition-specific expression analysis: Reveals experimental conditions under which recruited genes co-express.
• Functional module identification in genomics and systems biology: Extracts functionally relevant gene modules from microarray compendia for biological interpretation.

Methodology:

A Bayesian probabilistic model integrates priors from seed genes; a resolution sweep grows biclusters by recruiting genes with similar expression across subsets of conditions; the framework handles missing microarray values natively and has been validated on yeast compendia and artificial benchmark datasets.