Learning rule-based models of biological process from gene expression time profiles using gene ontology

Area of Science:

• Bioinformatics
• Computational Biology
• Systems Biology

Background:

• Microarray technology facilitates large-scale inference of gene participation in biological processes based on expression profiles.
• Current methods aim to develop models that automatically associate genes with biological processes using expression data and biological knowledge.

Purpose of the Study:

• To induce classificatory models from gene expression data and biological knowledge.
• To automatically associate genes with novel hypotheses of biological processes.

Main Methods:

• A systematic supervised learning approach is employed to predict biological processes from time-series gene expression data and biological knowledge.
• Gene Ontology is used to represent biological knowledge, which is then associated with expression-based features to form decision rules.
• The methodology is grounded in rough set theory.

Main Results:

• The developed rule model was evaluated using cross-validation on genes with known biological process roles.
• The model successfully generated hypotheses for genes with previously unknown biological process participation.
• The approach was demonstrated on a dataset published by Cho et al. (2001).

Conclusions:

• The study presents a robust method for predicting biological processes and generating novel gene function hypotheses.
• The integration of gene expression data and biological knowledge provides a powerful framework for functional genomics.
• The Rosetta system is available for further research and application.