Bioentity2vec

Bioentity2vec models bioentities by constructing a molecular association network that integrates genomic, chemical, and pathological data across 18 relationship types among eight bioentity classes to enable systematic molecular-level analysis.

Key Features:

• Molecular association network: Constructs a network that encapsulates 18 distinct relationships among eight bioentity classes.
• Integration of topological and biological information: Aggregates rich topological and biological information into the network representation.
• Aggregation and landscape discovery: Aggregates various bioentities to reveal physical and functional landscapes within biological systems.
• Machine learning-based evaluation: Employs machine learning techniques and uses a random forest classifier to evaluate relationships among bioentities.
• Predictive performance: Reports predictive metrics for the 18 relationships with an AUC of 0.9608 and a PR AUC of 0.9572.
• Multi-type relationship prediction: Supports simultaneous prediction of relationships involving single bioentity types and multiple bioentity types.
• Discriminative representations: Produces distinguishable representations that are beneficial for downstream classification tasks.

Scientific Applications:

• Relationship prediction: Predicts and evaluates associations among bioentities across multiple relationship types.
• Classification of bioentities and interactions: Provides discriminative features for classification tasks involving bioentities and their interactions.
• Molecular-level systems analysis: Enables systematic understanding of physical and functional landscapes at the molecular level.
• Support for experimental and industrial research: Informs experimental research and industrial product development through network-based insights.

Methodology:

Constructs a molecular association network integrating topological and biological information for 18 relationship types among eight bioentity classes and applies a random forest classifier to evaluate those relationships (AUC 0.9608; PR AUC 0.9572).