Gaussian Process Modeling of Protein Turnover

Area of Science:

• Biochemistry
• Systems Biology
• Computational Biology

Background:

• Protein turnover is crucial for cellular function and homeostasis.
• Existing nonstochastic models for protein turnover rate estimation have limitations.
• Stable-isotope labeling coupled with mass spectrometry is a powerful tool for studying protein dynamics.

Purpose of the Study:

• To develop and validate a novel stochastic model for calculating in vivo protein turnover rate constants.
• To compare the performance of the stochastic model against traditional nonstochastic methods.
• To investigate differences in protein turnover between mouse brain and liver tissues.

Main Methods:

• Development of a stochastic model based on stochastic differential equations, resulting in a Gaussian process with Ornstein-Uhlenbeck covariance.
• Application of the model to large-scale (15)N labeling data from mouse tissues.
• Validation using heavy water labeling in rat heart failure models, demonstrating isotope independence.

Main Results:

• The stochastic model provided significantly better fits to experimental data (99% of proteins) compared to nonstochastic curve fitting.
• Stochastic modeling revealed a 4-fold increase in the ratio of median degradation rate constants between liver and brain proteins compared to two-exponent fitting.
• The model predicted more pronounced differences in protein turnover between mouse liver and brain than previously estimated.

Conclusions:

• The developed stochastic model offers a more accurate and robust method for quantifying protein turnover rates.
• This approach enhances the understanding of tissue-specific protein dynamics and disease states.
• The model's isotope independence and freely available R implementation facilitate broader application in biological research.