Related Experiment Videos
Towards reconstruction of gene networks from expression data by supervised learning
Lev A Soinov1, Maria A Krestyaninova, Alvis Brazma
1Microarray Informatics Group, European Bioinformatics Institute, Wellcome Trust Genome Campus, Hinxton, Cambridge CB10 1SD, UK. lev@ebi.ac.uk
Genome Biology
|January 24, 2003
Summary
This study introduces a supervised learning method for gene network reconstruction. The approach effectively predicts gene expression, confirming known relationships and generating new hypotheses for cell-cycle gene interactions.
Area of Science:
- Computational Biology
- Systems Biology
- Bioinformatics
Background:
- Microarray experiments generate vast datasets crucial for gene network reconstruction.
- Identifying gene regulatory relationships (which genes affect others and how) is a key challenge.
Purpose of the Study:
- To develop a supervised learning approach for predicting gene expression based on other genes' expression data.
- To reconstruct gene networks by identifying causal relationships.
Main Methods:
- Utilized decision-tree-related classifiers for predicting gene expression.
- Developed algorithms applicable to continuous expression levels without requiring prior discretization.
- Applied the method to publicly available budding yeast cell cycle data.
Main Results:
- The method generated simple rules defining gene interrelations.
- Extracted rules largely confirmed existing knowledge of cell-cycle gene expression.
- Identified novel, previously unknown gene relationships as potential hypotheses.
Conclusions:
- The derived gene relationships are consistent with established scientific literature.
- The presented approach is validated as a reliable method for gene network reconstruction.
- The resulting rules can be utilized for building and explaining complex gene networks.