PredictiveNetworks

PredictiveNetworks infers gene interaction networks and evaluates their predictive power for phenotypes by integrating known gene interactions from biomedical literature and structured biological databases with gene expression data.

Key Features:

• Integration of Multiple Data Types: Incorporates known gene interactions from biomedical literature and structured biological databases together with gene expression data for network construction.
• Regression-based Network Inference: Uses a regression-based network inference algorithm to construct gene interaction graphs.
• Graph Topology: Constructs graphs that may include cycles but explicitly exclude self-loops.
• Predictive Modeling per Gene: Fits a prediction model for each gene based on its parent genes to assess the network's predictive ability.
• Analytical Pipeline Steps: Collects comprehensive sets of known gene interactions from publicly available sources and uses these interactions combined with gene expression data to infer networks.

Scientific Applications:

• Gene-list evaluation: Assess experimentally derived gene lists within the context of large-scale gene interaction networks.
• Phenotype and systems biology studies: Investigate genetic underpinnings of phenotypes and support genomics and systems biology research through network inference and predictive modeling.

Methodology:

Collect known gene interactions from biomedical literature and structured biological databases; combine with gene expression data; apply a regression-based network inference algorithm to construct graphs (cycles allowed, self-loops excluded); fit a prediction model for each gene using its parent genes to evaluate predictive ability.