Bayesian statistical analysis of circadian oscillations in fibroblasts

Area of Science:

• Biophysics
• Systems Biology
• Computational Biology

Background:

• Determining the period of biological oscillators from noisy, limited data is challenging.
• Current methods often provide only a point estimate, lacking uncertainty quantification.
• Reporting uncertainty measures is uncommon in experimental literature.

Purpose of the Study:

• Compare accuracy of six common period estimation methods.
• Introduce and evaluate a novel Bayesian method for period estimation.
• Assess the impact of cell number and sampling duration on uncertainty.

Main Methods:

• Comparative analysis of six period estimation techniques.
• Development and application of a Bayesian hierarchical model.
• Analysis of simulated data and experimental circadian gene expression data from mouse fibroblasts.

Main Results:

• The Bayesian method demonstrated superior accuracy in point estimates compared to six other methods.
• The Bayesian approach successfully quantified uncertainty in period estimates.
• Analysis revealed that substantial cell numbers may be needed to reduce uncertainty due to intrinsic cellular variability.

Conclusions:

• A Bayesian hierarchical model offers improved accuracy and uncertainty quantification for biological oscillator period estimation.
• Stochastic variability in intracellular oscillators necessitates careful consideration of experimental design (cell number, sampling duration).
• The model effectively deconvolutes within-cell stochasticity from population-level period variation.