Novel recurrent neural network for modelling biological networks: oscillatory p53 interaction dynamics

Area of Science:

• Systems Biology
• Computational Biology
• Biophysics

Background:

• Systems Biology research focuses on understanding cellular networks, gene, and protein interactions.
• Ordinary differential equations (ODEs) are common for modeling network dynamics but face challenges in parameter estimation and scalability.
• Limitations of ODE models hinder accurate temporal dynamics analysis in large biological systems.

Purpose of the Study:

• Introduce a novel recurrent artificial neural network (RNN) to model cellular signaling networks.
• Address limitations of ODE models, including parameter estimation and scalability for large systems.
• Quantify temporal dynamics and emergent system properties using a continuous model.

Main Methods:

• Developed a novel recurrent artificial neural network (RNN) based on ordinary differential equations (ODEs).
• Each neuron in the RNN represents the concentration change of a molecule, with weights corresponding to kinetic parameters.
• Applied the RNN to the p53-Mdm2 oscillation system, a key component of DNA damage response.

Main Results:

• The proposed RNN successfully represents the behavior of the p53-Mdm2 oscillation system.
• Achieved high accuracy in parameter estimation and captured system dynamics from sparse data.
• Investigated system robustness under parameter perturbations, yielding results consistent with biological knowledge.

Conclusions:

• The novel RNN offers an effective alternative to ODEs for modeling complex biological networks.
• The RNN facilitates accurate parameter estimation, temporal dynamics quantification, and robustness analysis.
• This approach holds promise for refining models as more quantitative data become available and for modularizing large signaling networks.