PANR

PANR infers functional gene interaction networks and modules from phenotypic responses to single-gene perturbations using a Bayesian mixture modeling framework.

Key Features:

• Posterior Association Networks (PANs): Represent posterior probabilities of pairwise gene associations inferred from phenotypic responses to single-gene perturbations and used to predict functional modules.
• Bayesian Mixture Modeling: Employs Bayesian mixture modeling to estimate gene associations and incorporate prior knowledge into association estimates.
• Hierarchical Clustering with Multiscale Bootstrap Resampling: Applies hierarchical clustering complemented by multiscale bootstrap resampling to identify statistically supported gene modules across scales.
• Scalability and Efficiency: Reduces the search space for combinatorial gene perturbation studies in large genomes (e.g., human) to improve the scalability of interaction discovery.

Scientific Applications:

• Ewing's Sarcoma: Identified a gene module containing confirmed and candidate therapeutic targets that are overrepresented in signaling pathways implicated in Ewing's sarcoma cell proliferation.
• Human Adult Stem Cells: Predicted a functional network of chromatin factors influencing epidermal stem cell fate, with validation by ChIP-seq, ChIP-qPCR, and RT-qPCR suggesting transcriptional cross-regulation among factors.

Methodology:

Integrates phenotypic data from single-gene perturbations, applies Bayesian mixture modeling to estimate pairwise gene associations and construct Posterior Association Networks, and uses hierarchical clustering with multiscale bootstrap resampling to detect significant modules while narrowing the combinatorial search space.