HPiP

HPiP predicts host-pathogen protein-protein interactions (PPIs) from amino acid sequences to identify molecular interfaces between pathogens and host proteins.

Key Features:

• Ensemble machine learning: Integrates multiple classifiers to improve prediction accuracy and robustness for PPI inference.
• Amino acid sequence property descriptors: Uses sequence-derived property descriptors as input features for predictive models.
• Validated on viral and bacterial datasets: Trained and evaluated on PPI sets including SARS-CoV-1, SARS-CoV-2 and human proteins, and applied to Mycobacterium tuberculosis.
• Experimental validation: Computational predictions have been experimentally confirmed using human monocyte THP-1 cells.
• Quality control metrics: Incorporates multiple QC metrics to assess prediction accuracy and reliability.

Scientific Applications:

• Systems-level mapping of host-pathogen interactions: Predicts previously unmapped PPIs to support systems biology analyses of infection.
• Identification of therapeutic targets: Helps identify host or pathogen proteins that may serve as targets for intervention.
• Drug repurposing support: Aids efforts to prioritize existing compounds by linking drugs to predicted host–pathogen interfaces.
• Cross-pathogen comparative studies: Enables comparative prediction across pathogens including SARS-CoV-1, SARS-CoV-2, and Mycobacterium tuberculosis.

Methodology:

HPiP extracts amino acid sequence property descriptors as sequence-based features and processes them through an ensemble of machine learning classifiers, integrating diverse feature types and combining multiple models' strengths.