Modelling HR-MAS NMR spectra of human cerebral organoids using COLMAR1D Deep Neural Networks – comparison against peak integration

Motivation: High-resolution magic-angle spinning (HR-MAS) NMR enables the study of metabolite concentrations in human cerebral organoids (COs). However, quantification remains challenging because HR-MAS spectra require precise NMR parameters for accurate modelling. Goal(s): Our goal is to assess the performance of a deep neural network-based deconvolution method to model CO HR-MAS spectra for metabolite quantification. Approach: CO HR-MAS spectra were modelled using the COLMAR1D Deep Neural Network Peak Picking tool. Quantification from COLMAR1D was compared against a peak integration. Results: 21 metabolites were quantified in cerebral organoids. A significant correlation between COLMAR1D and peak integration was observed, although COLMAR1D estimates were significantly higher. Impact: This study presents a convenient method for modelling and quantification of HR-MAS spectra in cerebral organoids using deep deconvolution. By reliably quantifying metabolite concentrations in human brain models, our approach will enable studies to better understand human brain metabolism.