Causal network inference using biochemical kinetics.

Area of Science:

• Biochemistry
• Systems Biology
• Computational Biology
• Network Inference

Background:

• Biochemical networks are crucial for understanding biological systems.
• Current network inference methods often rely on linear or discrete models, which may not capture complex non-linear dynamics.
• Non-linear formulations are hypothesized to improve causal network inference and prediction.

Purpose of the Study:

• To develop a general framework for network inference and dynamical prediction using non-linear biochemical kinetics.
• To infer both the network structure (chemical reaction graph) and kinetic parameters from time-course data.
• To enable prediction of system dynamics even when the underlying reaction graph is unknown or uncertain.

Main Methods:

• A Bayesian framework is employed for inference of the dynamical system, chemical reaction graph, and kinetic parameters.
• The framework utilizes time-course data from biochemical experiments.
• The approach is validated using simulated data from a mitogen-activated protein kinase signaling model and real phosphoproteomic data from cancer cell lines.

Main Results:

• The proposed non-linear framework demonstrates improved causal network inference compared to existing methods.
• The framework successfully predicts dynamical behavior and quantifies uncertainty, even when the reaction graph is initially unknown.
• Validation on both simulated and experimental data confirms the efficacy of the non-linear approach.

Conclusions:

• Non-linear biochemical kinetics provide a more accurate basis for network inference and dynamical prediction.
• The developed Bayesian framework offers a robust method for uncovering complex biochemical network structures and dynamics.
• This approach has significant implications for understanding and predicting the behavior of biological systems, particularly in disease contexts.