capDSD

capDSD extends the Diffusion State Distance (DSD) metric to integrate confidence measures, directed edges, and pathway data for enhanced functional inference in protein-protein interaction (PPI) networks.

Key Features:

• Incorporation of Network Noise Models: Accounts for network noise and uncertainty in PPI data to refine distance-based functional predictions.
• Confidence-weighted Edges (cDSD): Integrates edge confidence measures into the graph representation to modulate diffusion-based distances.
• Directional Edge Integration (caDSD): Incorporates known directed interactions to represent asymmetric relationships between proteins.
• Pathway Data Utilization (capDSD): Treats pathways as probabilistic units and integrates pathway information into the augmented graph representation.
• Incremental DSD Extensions: Builds on successive extensions—cDSD, caDSD, and capDSD—to progressively capture confidence, directionality, and pathway structure.

Scientific Applications:

• Protein function prediction: Employed in function prediction workflows, including a weighted majority vote method, to infer protein functions from PPI network structure.
• Benchmarking on yeast and STRING: Applied to the Baker's yeast PPI network and integrated protein association edges from the STRING database, where capDSD improved predictive performance relative to other matrices.

Methodology:

Calculates diffusion state distances on an augmented graph that includes confidence levels, directed interactions, and pathway data, using a diffusion-based metric to measure dissimilarity between protein node pairs.