Skip to main content
eScholarship
Open Access Publications from the University of California

UCLA

UCLA Electronic Theses and Dissertations bannerUCLA

Reducing Parameter Count in a Stochastic Model for Hair-Bundle Oscillations

Abstract

We explore how to reduce the number of parameters in biophysical models, which often sufferfrom an excess of parameters. Specifically, we simplify and fit a stochastic model for the active motility of hair bundles, constructed from experimental observations. Using statistical techniques from sensitivity analysis, we rank parameters in the model quantitatively. We then fix the least influential parameters, yielding a reduced deterministic model, validated using information criteria to ensure optimal predictive power. To fit a stochastic version of this model to measured hair bundles, we develop a suitable cost function to fit oscillatory models, applying it to the stochastic hair-bundle model to constrain its parameter values. Not only does this work contribute a general framework to fit stochastic oscillatory models to measured traces, but it also illuminates the key biophysical mechanisms in the model.