Nested sampling for parameter inference in systems biology: application to an exemplar circadian model.

Area of Science:

• Systems biology
• Computational biology
• Statistical modeling

Background:

• Model selection and parameter inference remain significant challenges in systems biology.
• Parameter inference provides parameter means and standard deviations, offering insights into data and model constraints.

Purpose of the Study:

• To apply nested sampling for parameter inference and Bayesian evidence estimation in a circadian rhythm model.
• To assess the impact of data length and sampling frequency on parameter uncertainty.

Main Methods:

• Utilized nested sampling, a Bayesian approach for model evidence calculation.
• Developed novel algorithms for likelihood calculation and nested sampling performance characterization.
• Applied methods to an established model of circadian rhythms.

Main Results:

• Demonstrated a ten-fold difference in the coefficient of variation between degradation and transcription parameters.
• Showed that parameter uncertainty decreases with increased data length (up to 4 cycles) but is unaffected by sampling frequency.
• Achieved significant improvements in computational efficiency for likelihood calculation and nested sampling.

Conclusions:

• Posterior parameter densities in circadian models are primarily influenced by time series length.
• Increasing the number of circadian cycles analyzed leads to more constrained parameter estimates.
• The coefficient of variation is a useful metric for distinguishing between well-constrained and less-constrained parameters.