Metabolites Identification and Quantification from 1H NMR Spectra by Database-Supported Bayesian Model Selection

• Activity Number: 582
• Type: Contributed
• Date/Time: Wednesday, August 4, 2010 : 2:00 PM to 3:50 PM
• Sponsor: Biometrics Section
• Author(s): Cheng Zheng*+ and Shucha Zhang and Daniel Raftery and Olga Vitek and Susanne Ragg
• Companies: Purdue University and Purdue University and Purdue University and Purdue University and Indiana University School of Medicine
• Keywords: Metabolomics ; Nuclear magnetic resonance (NMR) ; Metabolite identification ; Metabolite quantification ; Bayesian variable selection ; Mixed effects model

Abstract

1H NMR in metabolomics is a one-dimensional spectroscopy approach, used to measure the abundance of metabolites in mixtures. The peaks are often in complex shapes and heavily overlapped. Both identification and quantification of metabolites are challenging. To our knowledge, there is no automatic procedure for this task. We present a Bayesian database-based approach for simultaneously semiparametric baseline estimation, identifying and quantifying metabolites. We specify a Bayesian linear mixed model, and apply stochastic variable selection for both fixed and random effects. The approach preserved constraints on the parameter space imposed by the properties of NMR spectra. An efficient Gibbs sampling scheme is carried out to obtain models with high posterior probability. Our algorithm is demonstrated on both simulated NMR spectra and a real dataset from a coronary artery disease study.