Bayesian modeling and computation for analyte quantification in complex mixtures using Raman spectroscopy

Abstract

A two-stage algorithm based on Bayesian modeling and computation for quantifying analyte concentration in complex mixtures with Raman spectroscopy is proposed. A hierarchical Bayesian model is constructed for spectral signal analysis, and reversible-jump Markov chain Monte Carlo (RJMCMC) computation is carried out for model selection and spectral variable estimation. Processing is performed in two stages. In the first stage, the peak representation for a target analyte spectrum is learned. In the second, the peak variables learned from the first stage are used to estimate the concentration of the target analyte in a mixture. Numerical experiments validated the performance over a wide range of simulation conditions and established the algorithm accuracy over conventional multivariate regression algorithms for analyte quantification (when constrained to a small training sample size). In addition, the algorithm was applied to analyze experimental spontaneous Raman spectroscopy data collected for glucose concentration estimation in a biopharmaceutical process monitoring application. The results show that this algorithm can be a promising complementary tool alongside conventional multivariate regression algorithms in Raman spectroscopy-based mixture quantification studies, especially when collection of a large training dataset is challenging or resource-intensive.