Hierarchical Bayesian inference for ion channel screening dose-response data.

Area of Science:

• Pharmacology
• Computational Biology
• Biophysics

Background:

• Dose-response relationships are fundamental in biology and pharmacology.
• Hill curves are commonly used to model these relationships, typically reporting only best-fit parameters like IC50 and Hill coefficient.
• Existing methods often lack robust uncertainty quantification.

Purpose of the Study:

• Introduce PyHillFit, a novel Python software tool for dose-response curve analysis.
• Implement Bayesian inference methods to derive probability distributions for Hill curve parameters (IC50, Hill coefficient) and experimental noise.
• Enable hierarchical fitting to account for inter-experiment variability.

Main Methods:

• Utilized Bayesian inference for parameter estimation.
• Developed a Python-based software tool, PyHillFit.
• Applied hierarchical fitting to model inter-experiment variability.
• Compared maximum likelihood, Bayesian, and hierarchical Bayesian approaches.

Main Results:

• PyHillFit successfully infers probability distributions for dose-response parameters and noise levels.
• Hierarchical fitting effectively characterizes inter-experiment variability.
• Demonstrated application on a dataset of ion channel inhibition by drug compounds.
• Showcased uncertainty propagation into cardiac action potential simulations.

Conclusions:

• PyHillFit provides a robust framework for analyzing dose-response data with uncertainty quantification.
• Bayesian and hierarchical Bayesian approaches offer advantages over traditional methods.
• The tool facilitates more reliable predictions in pharmacological and biological modeling.