(149w) Parameter Estimation in Bioprocesses Using Bayesian Inference

This project provides a proof of concept for the parameter estimation for bioprocess based on rigorous and reliable confidence intervals. Bayesian Inference is used to estimate the uncertainty in the prediction of a parameter due to the presence of measurement noise in the process. While the major goal in this project is parameter estimation, one significant challenge in using Bayesian inference is the estimation of the evidence. In this regard, an algorithm called Nested Sampling is used to solve this problem by treating the evidence as an integral over the prior volume. The Nested Sampling algorithm works by continuously sampling from a prior distribution and calculating the likelihood of the model being “true”. Another issue previously associated with this algorithm was in the estimation of the likelihood itself. A high-fidelity mechanistic model typically uses ODE solvers such as RK45 which are relatively slow in solving and Nested Sampling usually converges after computing the model thousands of times. Therefore, a universal approximator such as a parameterized neural network is used in order to provide near instantaneous predictions making this technique feasible.

References:

J. Skilling, ‘Nested sampling for general Bayesian computation’, Bayesian Analysis, vol. 1, no. 4, pp. 833–859, 2006.