Integration of molecular network data reconstructs Gene Ontology

Area of Science:

• Computational Biology
• Bioinformatics
• Systems Biology

Background:

• Gene Ontology (GO) evaluation is shifting towards using molecular network data for its construction and assessment.
• Previous work demonstrated that GO can be largely reconstructed from molecular network topologies alone.
• A need exists for integrating multiple molecular network types to reconstruct and update GO.

Purpose of the Study:

• To develop a novel data integration framework for reconstructing and updating Gene Ontology (GO).
• To assess the extent to which GO can be recovered by integrating various molecular interaction data.
• To infer and validate new relationships between genes and GO terms, and between GO terms themselves.

Main Methods:

• A computational framework using penalized non-negative matrix tri-factorization (PNMTF) for integrating diverse biological networks.
• Simultaneous clustering of genes and GO terms to identify novel relationships.
• Incorporation of additional topological information, specifically wiring similarity of non-interacting genes, to enhance accuracy.

Main Results:

• Integration of protein-protein interaction, genetic interaction (GI), and co-expression networks recovered 96% of directly related GO terms.
• The inclusion of wiring similarity improved GO term association capture by 6%.
• 44% of inferred GO term relationships were validated in the literature, with significant performance in Cellular Component, Molecular Function, and Biological Process domains.

Conclusions:

• The developed framework effectively integrates multiple molecular network data types for GO reconstruction and update.
• The method demonstrates high accuracy in recovering existing GO relationships and predicting novel ones.
• This approach advances the understanding and annotation of gene functions and biological pathways.