Data- and expert-driven rule induction and filtering framework for functional interpretation and description of gene...

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• High-throughput molecular biology generates vast experimental data requiring interpretation.
• Manual functional gene/protein group analysis is time-consuming and costly.
• Efficient bioinformatics tools are needed for functional analysis of experimental results.

Purpose of the Study:

• To develop a comprehensive framework for functional interpretation of gene sets using Gene Ontology (GO) terms.
• To present automated and expert-driven methods for rule induction and filtering.
• To support experts in analyzing large-scale biological data.

Main Methods:

• Developed a framework for inducing logical rules from GO term combinations.
• Implemented four approaches: automated rule induction (with/without filtering) and expert-driven methods (utility functions, seed terms).
• Compared different rule induction and filtering algorithms.

Main Results:

• Filtering reduces the number of rules for expert analysis but may omit relevant information.
• Expert-driven methods, especially using seed terms, effectively reduce rules while retaining essential information.
• The seed term method ensures all rules incorporate expert-defined terms.

Conclusions:

• A filtering step is crucial for managing rule sets in functional gene analysis.
• Expert interaction during rule induction enhances interpretability and retains valuable insights.
• The seed term approach facilitates discovering novel biological process combinations.
• A Matlab script suite is available for the presented framework.