RestoreNet

RestoreNet models clonal dynamics to quantify and investigate clonal dominance in gene therapy safety studies and haematopoiesis using mixed-effects stochastic frameworks.

Key Features:

• Mixed-effects stochastic models: Implements mixed-effects stochastic models to capture clone-level and population-level variability.
• High-throughput clonal tracking calibration: Calibrates stochastic differential equations from high-throughput clonal tracking data quantifying cells derived from individual hematopoietic stem cell ancestors.
• Stochastic reaction networks with mixed-effects GLMs: Integrates stochastic reaction networks with mixed-effects generalized linear models to analyze clonal dominance in high-dimensional clonal datasets.
• Kramers-Moyal approximated Master equation: Employs the Kramers-Moyal approximated Master equation for local linear approximation of duplication, death, and differentiation at the clonal level.
• Maximum likelihood estimation: Infers model parameters via maximum likelihood estimation.
• Random effects for clonal heterogeneity: Incorporates random effects on clonal parameters to model heterogeneity in fitness leading to dominance.
• Expectation-maximization calibration: Calibrates the random-effects extension using a tailor-made expectation-maximization algorithm.
• Simulation benchmarking: Demonstrates improved performance over state-of-the-art methods in simulation studies.
• In vivo analysis: Applied to two in‑vivo studies to uncover patterns of clonal dominance.

Scientific Applications:

• Gene therapy safety assessment: Quantifies abnormal clonal expansions following gene therapy to support safety analyses.
• Haematopoiesis research: Investigates hierarchical relationships and dynamics of clonal populations in haematopoiesis.
• Parameter inference for stochastic population models: Estimates duplication, death, and differentiation parameters from clonal tracking data.
• Method benchmarking with simulations and in vivo data: Compares model performance against existing methods using simulations and empirical in‑vivo datasets.

Methodology:

Calibrates stochastic differential equations from high-throughput clonal tracking data; integrates stochastic reaction networks with mixed-effects generalized linear models; uses the Kramers-Moyal approximated Master equation with local linear approximation; performs parameter inference via maximum likelihood estimation; and calibrates random effects with an expectation-maximization algorithm.