SSIF

SSIF identifies potentially missing "is-a" relationships in the Gene Ontology to improve hierarchical consistency and semantic accuracy.

Key Features:

• Automated ontology auditing: Automates detection of potentially missing "is-a" relationships within the Gene Ontology.
• Term-algebra model: Employs a novel term-algebra approach built on a sequence-based representation of GO concepts.
• Conditional inference rules: Applies three conditional rules—monotonicity, intersection, and sub-concept—to systematically generate candidate "is-a" relations.
• Dataset application: Applied to the Gene Ontology release from 2018-10-03 and identified 1,938 unique potentially missing "is-a" relations.
• Expert validation results: Expert review of a random sample of 210 suggested relations yielded precision of 60.61% (monotonicity), 60.49% (intersection), and 46.03% (sub-concept).
• Implementation: Implemented in Java.

Scientific Applications:

• Gene Ontology quality control: Detects candidate missing "is-a" links to support improvement of GO hierarchical integrity and semantic consistency.
• Ontology curation support: Generates candidate relations for expert curators to review and incorporate into GO.
• Downstream data consistency: Helps reduce semantic errors that could affect analyses relying on GO annotations.

Methodology:

Uses a sequence-based representation of GO concepts and a term-algebra approach that applies three conditional rules (monotonicity, intersection, sub-concept) to automatically suggest potentially missing "is-a" relationships; applied to the Gene Ontology release from 2018-10-03.