Predicting the protein half-life in tissue from its cellular properties

Area of Science:

• Proteomics and Systems Biology
• Molecular Biology and Biochemistry

Background:

• Protein half-life is crucial for maintaining protein homeostasis (proteostasis).
• High-throughput proteomics studies provide protein half-life estimates in both tissues and cells.
• Significant differences exist between cellular and tissue protein half-lives, with more data available from cell studies.

Purpose of the Study:

• To develop a multivariate linear model for predicting tissue protein half-life using cellular properties.
• To identify and leverage substructures within protein half-life data by clustering.
• To improve the accuracy of tissue protein half-life predictions by accounting for correlations between cell and tissue data.

Main Methods:

• Designed a multivariate linear model incorporating cellular half-life, abundance, intrinsically disordered sequences, and transcriptional/translational rates.
• Utilized relative distance from regression lines to identify three distinct clusters in tissue vs. cell protein half-life data.
• Trained and applied the model to predict protein half-lives in murine liver, brain, and heart tissues.

Main Results:

• Observed similar prediction patterns across different murine tissue types (liver, brain, heart).
• Found that model performance is strongest when tissue and cell culture protein half-lives are highly correlated.
• Clustering identified variations in correlation coefficients, improving overall prediction and revealing distinct subgroups of proteins.

Conclusions:

• The developed model effectively predicts tissue protein half-lives from cellular data, offering a generalizable approach.
• Clustering protein half-lives enhances predictive accuracy by accounting for varying correlations between cell and tissue data.
• The model and its implementation in R code provide a valuable tool for researchers studying protein homeostasis across different biological contexts.