Stochastically modeling multiscale stationary biological processes.

Area of Science:

• Computational Biology
• Systems Biology
• Toxicology

Background:

• Biological systems exhibit inherent uncertainty due to noise and measurement errors, complicating computational modeling.
• Existing models like multiscale stochastic models, chemical kinetic models, and stochastic differential equations have limitations in speed, noise capture, or biological fidelity.

Purpose of the Study:

• To develop a new computational approach for modeling multiscale stationary biological processes that incorporates experimental data noise.
• To estimate parameter uncertainties and identify potential model mis-specifications.

Main Methods:

• Modeling the mean stationary response at each biological level based on expected response relationships.
• Utilizing conditional Monte Carlo sampling to capture variations around the mean, ensuring statistical consistency with training data.
• Reconstructing conditional probability distributions for biological responses to address parameter identification challenges.

Main Results:

• The novel approach successfully models biological system noise and provides estimates for parameter uncertainties.
• It overcomes the parameter identification problem by reconstructing conditional probability distributions.
• Demonstrated application in modeling variations across multiple organizational scales within a teleost reproduction-related Adverse Outcome Pathway.

Conclusions:

• The developed method offers a robust way to model uncertain biological responses, complementing existing dynamical modeling techniques.
• It enhances the prediction of biological responses over experimental time scales by embracing data noise.
• The approach has practical applications in toxicological studies, specifically in understanding Adverse Outcome Pathways.