A critical step in the analysis of many 1D, 2D, and higher dimensional NMR spectra is the accurate identification of peaks, which is often followed by full quantitation. Current approaches typically require manual intervention considerably slowing down the workflow. We recently developed the deep neural network “DEEP Picker” for the automated analysis of crowded spectra of proteins and complex mixtures of small molecules. I will demonstrate the performance of DEEP Picker for 1D, 2D, and pseudo-3D data of proteins and metabolomics mixtures and how it can be combined with our “Voigt Fitter” software for full quantitation of peak positions, linewidths, and volumes. We implemented these new tools in the public web server COLMARq (http://spin.ccic.osu.edu/index.php/quan/index) for the automated analysis of cohorts of metabolomics samples for metabolite identification, quantitation, and statistical analysis. I will discuss the new methodology with examples with proteins and metabolomics for the identification of metabolic biomarkers in synovial fluid infected with Pseudomonas aeruginosa, an opportunistic pathogen that has the ability to form biofilm.
