Principled Decision-Making Workflow with Hierarchical Bayesian Models of High-Throughput Dose-Response Measurements

Area of Science:

• Biophysics
• Statistical Modeling
• High-Throughput Screening

Background:

• High-throughput protein melting-point assays generate large datasets.
• Data quality can be compromised by noise, complicating analysis.
• Accurate determination of protein stability is crucial for various biological applications.

Purpose of the Study:

• To apply hierarchical Bayesian estimation to high-throughput protein melting-point data.
• To assess the model's ability to handle noisy data and impute melting temperatures.
• To demonstrate a principled decision-making framework using posterior distribution variance.

Main Methods:

• Hierarchical Bayesian estimation framework.
• Analysis of protein melting-point data across diverse organisms.
• Comparison with maximum-likelihood curve-fitting methods.

Main Results:

• The Bayesian model successfully imputed reasonable melting temperatures despite significant data noise.
• Variance in posterior distribution estimates provided a basis for informed experimental decisions.
• The proposed workflow offered advantages over standard curve-fitting approaches.

Conclusions:

• Hierarchical Bayesian estimation is a robust method for analyzing noisy high-throughput protein stability data.
• Utilizing posterior distribution variance enables more principled decision-making in experimental contexts.
• This approach offers a valuable alternative to traditional curve-fitting techniques for optimizing high-throughput measurements.