A Framework for Parameter Estimation and Uncertainty Quantification in Systems Biology Using Quantile Regression and...

Area of Science:

• Computational Biology
• Systems Biology
• Machine Learning

Background:

• Accurate parameter estimation and uncertainty quantification are vital for understanding complex biological systems.
• Current methods face challenges in achieving high accuracy at reasonable computational costs.

Purpose of the Study:

• To develop a novel, efficient framework for parameter estimation and uncertainty quantification in systems biology.
• To integrate the quantile method with Physics-Informed Neural Networks (PINNs) for improved biological modeling.

Main Methods:

• Developed a novel approach integrating the quantile method with Physics-Informed Neural Networks (PINNs).
• Utilized a multi-output neural network architecture to characterize parameter estimation and uncertainty.
• Validated the approach across three case studies and a larger-scale model, comparing it with Monte Carlo dropout (MCD) and Bayesian methods.

Main Results:

• The proposed approach demonstrated significantly superior efficacy in parameter estimation and uncertainty quantification compared to MCD and Bayesian methods.
• Achieved high accuracy in characterizing parameter estimation and associated uncertainty.
• Showcased excellent performance on a larger-scale biological model.

Conclusions:

• The novel quantile method and PINNs integration offers a powerful tool for systems biology modeling.
• This approach promises to expand the applications of computational modeling in biological research.
• Provides a more accurate and computationally efficient solution for parameter estimation and uncertainty quantification.