ASIAN

ASIAN infers gene regulatory networks from large-scale gene expression profiles using hierarchical clustering, multiple regression with variance inflation factors, and graphical Gaussian modeling (GGM).

Key Features:

• Integration of analytical techniques: Combines hierarchical clustering, multiple regression (using variance inflation factors), and graphical Gaussian modeling (GGM) to identify gene clusters and infer partial-correlation networks.
• Automated cluster analysis: Determines cluster boundaries automatically by applying variance inflation factors within multiple regression to assess multicollinearity and guide clustering decisions.
• Graphical Gaussian Modeling (GGM): Computes the inverse of the correlation coefficient matrix to estimate partial correlations and infer putative regulatory interactions among clusters.
• Regression-based multicollinearity assessment: Uses variance inflation factors in multiple regression to quantify multicollinearity among expression variables for more robust cluster definition.
• Output generation: Produces dendrograms, estimated cluster numbers, and graphical representations of inferred networks as analytical outputs.
• Scalability and efficiency: Employs a mathematical framework intended to process large gene expression datasets within practical computational times.

Scientific Applications:

• Inference of gene regulatory networks: Reconstruction of putative regulatory relationships among genes or gene clusters from expression data.
• Functional genomics studies: Clustering of genes by expression profiles to support functional annotation and hypothesis generation about gene function.
• Comparative genomic analysis: Comparison of inferred regulatory networks across conditions or species to investigate evolutionary or disease-related differences.

Methodology:

Hierarchical clustering of gene expression data precedes automated cluster-boundary determination using variance inflation factors within multiple regression; GGM is then applied by inverting the correlation coefficient matrix to estimate partial correlations between clusters and construct inferred networks, with outputs including dendrograms, estimated cluster numbers, and graphical network representations.