MINER: exploratory analysis of gene interaction networks by machine learning from expression data.

Area of Science:

• Systems Biology
• Bioinformatics
• Computational Biology

Background:

• Reconstructing gene regulatory networks from high-throughput omics data is crucial for modeling living systems.
• Existing methods often perform "one-shot" network reconstruction or simple correlation analyses.
• There is a need for interactive tools that incorporate biological domain knowledge.

Purpose of the Study:

• To develop an interactive application for exploring gene regulatory dependencies.
• To enable biologists to formulate hypotheses and discover novel regulatory relationships.
• To visualize gene dependencies using both tree and network representations.

Main Methods:

• Developed MINER (Microarray Interactive Network Exploration and Representation) application.
• Employs multivariate non-linear tree learning for individual gene dependencies.
• Integrates visualization of dependencies as trees and networks.
• Incorporates known biological relationships via Gene Ontology annotations.

Main Results:

• MINER allows interactive exploration of gene expression data.
• Biologists can guide network exploration using domain knowledge.
• Multiple trees can be summarized into gene network diagrams.
• MINER has led to the discovery of a novel, experimentally validated regulatory relationship.

Conclusions:

• MINER facilitates gene-by-gene network construction, incorporating user expertise.
• This approach differs from most automated gene regulatory network inference methods.
• Applicable to RNA microarray, proteomics, and next-generation sequencing data.