ternarynet

ternarynet estimates ternary gene regulatory networks from gene perturbation experiment data using a computational Bayesian approach to infer network structure and steady states.

Key Features:

• Implicit Modeling Approach: Models persistent perturbation effects on a gene network, predicts the steady state of the modified network, and compares predicted states with observed experimental data to infer the original unperturbed network.
• Bayesian Framework: Uses Bayesian inference to address the many-to-one inverse problem and to quantify posterior probabilities of network structures and steady-state behaviors.
• Model Uncertainty Resolution: Identifies uncertainties in model features and guides additional perturbation experiments to resolve ambiguities in network reconstruction.
• Consistency and Convergence: Applies randomized fittings of the algorithm that converge toward a common posterior density on the space of models, providing robustness beyond single-model outputs.
• Application to Oncogenic Mutations: Has been applied to model gene regulatory networks involving genes responsive to oncogenic mutations, generating hypothetical models consistent with known regulatory properties.

Scientific Applications:

• Gene regulatory network reconstruction: Reconstruction of ternary gene regulatory networks from perturbation data and steady-state observations.
• Oncogenic mutation analysis: Modeling gene responses and regulatory interactions in the context of oncogenic mutations and synergistic gene responses.
• Experimental design for perturbation studies: Informing selection and interpretation of additional perturbation experiments to reduce model uncertainty and improve inference.

Methodology:

Implicitly model persistent perturbation effects, predict the steady state of the perturbed network, and compare predictions to observed experimental data; employ Bayesian inference to resolve the many-to-one inverse problem and quantify posterior probabilities; use randomized applications of the fitting algorithm that converge toward a common posterior density and iteratively resolve uncertainties via additional perturbations.