GENetRank

GENetRank prioritizes genes within protein-protein interaction networks to identify regulators of drug pharmacodynamics using asymmetric random walks with restarts, absorbing states, a renormalization scheme, and saturation indices.

Key Features:

• Asymmetric random walks with restarts and absorbing states: Employs asymmetric random walks incorporating restarts and absorbing states to traverse and analyze complex PPINs more progressively than traditional random walk methods.
• Renormalization scheme: Uses a novel renormalization scheme to adjust scores and improve exploration of PPIN topology.
• Saturation indices: Introduces saturation indices to quantify progressive exploration and to optimize the interaction between absorbing states and renormalization.
• Two-set gene prioritization: Prioritizes genes by contrasting a set predictive of drug sensitivity with a set involved in the drug's mechanism of action within the PPIN.
• Application to TRAIL single-cell data: Applied to a predictive gene signature of cancer cell sensitivity to TRAIL (tumor-necrosis-factor-related apoptosis-inducing ligand) using single-cell analyses.
• Enriched pharmacodynamics gene sets: Produces prioritized gene sets that are significantly enriched for genes regulating drug pharmacodynamics.
• Gene Expression Radars: Provides gene expression radar visualizations to assess all pairwise interactions within the network for comprehensive comparison.

Scientific Applications:

• Drug mechanism exploration: Identifies key regulatory genes to elucidate mechanisms of action of drugs.
• Molecular target discovery: Supports discovery of candidate molecular targets for co-treatment strategies by prioritizing genes linked to drug sensitivity.
• Pathway mining: Mines gene sets related to signaling pathways, offering more specific candidates when other methods yield large, nonspecific sets.

Methodology:

Defines two gene sets within a PPIN (one predictive of drug sensitivity and one involved in the drug's mechanism) and applies asymmetric random walks with restarts and absorbing states together with a renormalization scheme and saturation indices; validation was performed using the MINT network and the method was applied to single-cell analyses of TRAIL-sensitive cancer cells.