Designing a novel radial basis neural structure for solving the dynamical hepatitis C virus model

Area of Science:

• Computational biology
• Mathematical modeling
• Virology

Background:

• Hepatitis C virus (HCV) infection dynamics involve complex interactions between viruses, infected, and uninfected hepatocytes.
• Understanding these dynamics is crucial for developing effective treatment strategies, especially in patients with high baseline viral loads.
• Nonlinear dynamical systems are often employed to represent biological processes like viral infections.

Purpose of the Study:

• To design and implement a novel radial basis neural network (RBNN) for solving the dynamical hepatitis C virus model.
• To address the nonlinear dynamical structure of HCV infection in patients with high baseline viral loads.
• To optimize the RBNN model using a Bayesian regularization scheme for enhanced accuracy.

Main Methods:

• A feedforward neural network with radial basis functions and 20 neurons was utilized.
• The Bayesian regularization approach was employed for optimization.
• A reference solution was generated using the explicit Runge-Kutta method with a step size of 0.01.
• Data was divided into training (72%), validation (14%), and testing (14%) sets.

Main Results:

• The RBNN model demonstrated high precision in solving the HCV dynamical model.
• Achieved absolute error values were in the range of 10^-06 to 10^-08.
• Outcomes showed good overlapping, indicating the reliability of the proposed scheme.

Conclusions:

• The developed RBNN with Bayesian regularization is an effective and precise solver for the nonlinear dynamical hepatitis C virus model.
• The approach provides a robust tool for analyzing HCV infection dynamics, particularly in high viral load scenarios.
• Statistical evaluations confirmed the efficiency and accuracy of the proposed computational method.