ddgraph

ddgraph infers combinatorial regulatory codes among transcription factors from genome-wide TF binding profiles, distinguishing direct from indirect interactions despite high correlation and weak associations among features.

Key Features:

• Neighbourhood Consistent PC (NCPC): The NCPC algorithm handles highly correlated features, weak associations, and small sample sizes and distinguishes direct from indirect interactions by analyzing the dependence structure around target variables.
• Direct Dependence Graphs (DDGraph): DDGraph constructs direct dependence graphs that represent complex dependencies and direct associations among transcription factors and gene expression.
• Integration with Graphical Modelling: Integrates with graphical modelling approaches, including Bayesian Networks, to facilitate distinction between direct and indirect interactions.

Scientific Applications:

• Transcription Factor Analysis: Identifies transcription factors that directly regulate gene expression versus those indirectly associated with it.
• Cis-Regulatory Module (CRM) Classification: Applied to Drosophila mesoderm differentiation data to identify TFs specifying different CRM classes and to detect patterns such as depletion of Twist binding at CRMs regulating expression in specific muscle cells.

Methodology:

Uses graphical models implementing the NCPC algorithm to analyze dependence structures around target variables and constructs Direct Dependence Graphs (DDGraph), with interfaces to Bayesian network approaches.