PPHAGE

PPHAGE predicts and prioritizes human aging-associated genes genome-wide by integrating evidence from 11 distinct databases and applying Naïve Bayes classifiers together with positive-unlabeled learning methods (NB, Spy, Rocchio-SVM).

Key Features:

• Data Fusion: Integrates information from 11 distinct biological databases to provide multi-source evidence for gene aging associations.
• Machine Learning Algorithms: Uses Naïve Bayes classifiers alongside positive-unlabeled learning (PUL) methods including NB, Spy, and Rocchio-SVM to score and rank genes.
• Negative Sample Identification: Applies PUL methods to identify putative negative (non-aging) genes to complement known positive seed genes.
• Algorithm Fusion: Combines multiple PUL algorithms when no single method is superior, yielding a fused ranking approach for improved robustness.
• Candidate Gene Prediction and Prioritization: Produces a prioritized set of over 3,000 candidate age-related genes linked to biological pathways, ontologies, and diseases such as cancer and diabetes.

Scientific Applications:

• Candidate discovery: Generate prioritized lists of putative aging-associated genes for downstream experimental validation.
• Pathway and ontology mapping: Associate ranked candidate genes with biological pathways and Gene Ontology terms relevant to aging.
• Disease association analysis: Identify candidate genes connected to age-related diseases, exemplified by links to cancer and diabetes.
• Comparative algorithm assessment: Evaluate how different PUL algorithms contribute to negative sample identification and ranking stability.

Methodology:

Systematic fusion of multiple databases combined with Naïve Bayes classifiers and positive-unlabeled learning methods (NB, Spy, Rocchio-SVM), using PUL-identified negatives and an algorithm-fusion strategy to rank and prioritize candidate aging genes.