PALMER

PALMER identifies indirect gene–gene relationships from biomedical literature by applying a constrained latent block model that integrates Gene Ontology and KEGG prior knowledge to enhance pathway annotation.

Key Features:

• Constrained latent block model: Implements a constrained latent block (biclustering) model tailored for literature mining data.
• Integration of prior biological knowledge: Incorporates prior annotations from Gene Ontology and KEGG to guide the model.
• Identification of indirect gene relationships: Detects indirect associations between genes present in biomedical literature.
• Automatic GO-term association: Associates relevant Gene Ontology terms automatically with identified gene relationships to aid interpretation.
• KEGG-guided discovery: Leverages existing KEGG pathway annotations to inform pathway-related findings.
• Improved weak-signal performance: Demonstrates superior performance relative to traditional latent block models in scenarios with weak signals.
• Evaluation on simulations and real data: Validated through simulation studies and real-world applications on pathway-modulating genes.

Scientific Applications:

• Discovery of novel gene–gene associations: Mining biomedical literature to uncover previously unreported relationships between genes.
• Pathway annotation enhancement: Improving pathway annotation by integrating literature-derived associations with KEGG and Gene Ontology.
• Analysis of pathway-modulating genes in cancer: Applied to identify genes modulating cancer signaling pathways.
• Literature-based functional interpretation: Facilitating biological interpretation of literature mining results via automatic GO-term assignment.

Methodology:

Uses a constrained latent block model that integrates prior knowledge from Gene Ontology and KEGG and automatically associates Gene Ontology terms with identified gene relationships; performance was assessed via simulation studies and real-world applications on pathway-modulating genes in cancer signaling pathways.