Targeted retrieval of gene expression measurements using regulatory models

Area of Science:

• Bioinformatics
• Computational Biology
• Systems Biology

Background:

• Public gene expression repositories enable contextualizing new experiments with prior studies.
• Existing methods for experiment comparison often rely on experimental annotations or global expression similarities.
• A need exists for methods that leverage data-driven insights into gene regulatory networks for experiment retrieval.

Purpose of the Study:

• To develop a novel approach for comparing gene expression experiments using unsupervised, data-driven regulatory models.
• To improve the retrieval of relevant biological experiments by focusing on regulatory relationships of pre-specified genes.
• To provide a method for interpreting biological conditions through changes in inferred regulatory link activity.

Main Methods:

• Learning gene regulatory models from public gene expression data repositories.
• Employing the Fisher kernel to construct a similarity metric based on inferred regulatory links.
• Applying the method to human and plant microarray datasets for experiment retrieval.

Main Results:

• The proposed method significantly improves the retrieval of related experiments compared to standard approaches.
• The approach allows for the interpretation of biological conditions by analyzing changes in regulatory link patterns.
• Successful identification of relevant gene relationships in the osmotic stress network of Arabidopsis thaliana.

Conclusions:

• Data-driven regulatory models offer a powerful framework for comparing and retrieving gene expression experiments.
• The targeted, model-based similarity measure enhances interpretability and biological relevance in experiment retrieval.
• This method provides a valuable tool for researchers analyzing large-scale gene expression data.