Conceptualization of molecular findings by mining gene annotations

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• The Gene Ontology (GO) provides a framework for gene and gene product annotation.
• Current GO annotations are highly specific, making it difficult to identify major functional themes from large gene lists generated by genome-scale studies.
• There is a need for tools to objectively and quantitatively mine semantic information for conceptual-level functional theme discovery.

Purpose of the Study:

• To develop a method for deriving abstract representations of major biological processes from gene lists.
• To identify functionally coherent gene subsets summarized by informative GO terms.
• To objectively and quantitatively capture major gene function directions in a context-specific manner.

Main Methods:

• Utilized the hierarchical structure of the Gene Ontology (GO).
• Developed a method to identify non-disjoint, functionally coherent gene subsets.
• Employed information-content-based measures and graph-based statistics derived from Steiner trees for evaluation.

Main Results:

• Evaluated various metrics for information loss during GO term merging.
• Assessed different statistical schemes for evaluating gene set functional coherence.
• Identified a combination of information-content and graph-based statistics yielding the best discriminative power.

Conclusions:

• The developed methods offer an objective and quantitative approach to functional theme discovery.
• The approach effectively captures major gene function directions in a context-specific manner.
• Facilitates interpretation of large gene lists from genome-scale studies.