SemDist

SemDist computes semantic distances and similarities between predicted and true Gene Ontology (GO) protein function annotations to quantify remaining uncertainty and misinformation in genomic annotation datasets.

Key Features:

• Semantic distance and similarity calculation: Computes semantic distance and similarity scores between sets of predicted and true Gene Ontology (GO) annotations for protein function.
• Information accretion-based uncertainty and misinformation quantification: Uses information accretion methodologies to quantify remaining uncertainty, misinformation, and information gain across GO versions.
• Information-theoretic statistical framework: Implements a statistical framework based on principles of information theory to assess annotation quality.

Scientific Applications:

• Protein Function Prediction Evaluation: Validates and benchmarks protein function prediction models by comparing predicted annotations to true GO annotations using semantic distance and similarity metrics.
• Gene Ontology Analysis: Analyzes semantic relationships and information content across Gene Ontology versions to interpret annotation changes and information gain.

Methodology:

SemDist applies information accretion methods and a statistical framework based on information theory to compute semantic distances and to quantify the new information gained from specific versions of the Gene Ontology (GO) for assessing predicted versus true protein function annotations.