Parameter inference for a stochastic kinetic model of expanded polyglutamine proteins

Area of Science:

• Computational Biology
• Biophysics
• Disease Modeling

Background:

• Protein aggregates are hallmarks of age-related diseases, including Huntington's disease.
• Existing models of expanded polyglutamine (PolyQ) protein dynamics rely on fixed parameters with significant uncertainty.
• Current parameter estimation relies on ad hoc tuning to match limited experimental data.

Purpose of the Study:

• To develop robust inference procedures for estimating model parameters in the presence of uncertainty and limited data.
• To overcome the computational expense of traditional simulation-based inference methods.
• To provide new insights into the biological processes underlying PolyQ protein aggregation.

Main Methods:

• Utilized a computational model for the dynamic evolution of expanded polyglutamine (PolyQ) proteins.
• Developed Gaussian process emulators as fast stochastic approximations for key model quantities.
• Reformulated the inference algorithm around these computationally efficient emulators.

Main Results:

• Successfully constructed Gaussian process emulators to approximate computationally expensive simulations.
• Enabled a more efficient and accurate estimation of model parameters.
• Identified appropriate parameter values providing novel insights into PolyQ protein dynamics.

Conclusions:

• The developed methods offer a computationally efficient approach to parameter inference for complex biological models.
• Accurate parameter estimation is crucial for understanding disease mechanisms.
• The findings highlight key parameter values that advance our understanding of age-related protein aggregation diseases.