GOntoSim

GOntoSim quantifies functional similarity between genes by comparing Gene Ontology (GO) terms using the ontology graph structure and node information content.

Key Features:

• Pairwise GO-term similarity: Quantifies similarity between pairs of GO terms using both graph structure and node information content.
• Ancestor and descendant consideration: Measures similarities among direct ancestors of GO terms and their common descendants to capture extended functional relationships.
• Benchmark dataset evaluation: Evaluated on the Enzyme Dataset comprising 10,890 proteins and 97,544 GO annotations.
• Enzyme clustering and EC comparison: Clusters enzymes and compares clusters against Gold Standard EC numbers to assess functional grouping accuracy.
• Performance at EC level 1: Achieves a purity score of 0.75 at EC level 1, compared to 0.47 for GOGO and 0.51 for Wang.
• Robustness to IEA annotations: Maintains a purity score of 0.94 at EC level 1 when using annotations with the Inferred from Electronic Annotation (IEA) evidence code, versus 0.48 for both GOGO and Wang.

Scientific Applications:

• Data annotation: Supports transfer and refinement of GO-based functional annotations across genes and proteins.
• Knowledge transfer: Enables propagation of functional information between annotated and unannotated entities based on GO similarity.
• Identification of functionally related genes: Detects groups of genes or proteins with similar molecular functions using GO-term similarity.
• Enzyme functional classification: Assesses and refines enzyme groupings by comparison to EC number standards.

Methodology:

Compute pairwise GO-term similarity using the GO graph structure and node information content, including direct ancestors and common descendants; cluster enzymes and compare clusters to Gold Standard EC numbers; evaluation performed on the Enzyme Dataset of 10,890 proteins and 97,544 GO annotations.