Pigengene

Pigengene infers and evaluates biological signatures from gene expression profiles across different platforms to identify coexpression modules and build predictive models for disease.

Key Features:

• Cross-Platform Biological Signature Inference: Infers signatures on one gene expression platform and evaluates them on another to produce technology-independent signatures.
• Coexpression Network Analysis: Identifies clusters of highly coexpressed genes and summarizes each cluster by an eigengene representing module-level expression.
• Bayesian Network Modeling: Constructs a Bayesian network to model probabilistic dependencies among eigengene-represented modules.
• Decision Tree Construction: Builds decision trees based on eigengene expression profiles to predict diagnostic and prognostic outcomes.
• Cross-Dataset Training and Validation: Trains models on one dataset (e.g., microarray) and validates predictions on an independent dataset (e.g., RNA-Seq) to assess sensitivity and specificity.

Scientific Applications:

• Hematological malignancy discrimination: Delineates biological differences between myelodysplastic syndrome (MDS) and acute myeloid leukemia (AML) using eigengene expression.
• Pathway-specific gene identification: Identifies differentially expressed genes within pathways such as the extracellular matrix pathway, including underexpression in AML relative to MDS.
• Molecular validation: Supports findings with experimental validations including immunocytochemistry and methylation analysis, noting hypermethylation of MMP9 among AML cases.
• Generalizability to other diseases: Applies the network analysis approach to other complex diseases such as breast cancer prognosis to define disease-specific biological signatures.

Methodology:

Performs coexpression network analysis, summarizes modules by eigengenes, constructs Bayesian networks of eigengenes, builds eigengene-based decision trees, and trains on one dataset with validation on an independent dataset (e.g., microarray to RNA-Seq).