Fractal Approach for 1H-NMR Spectra Simplification and Data Processing

Abstract

Nuclear magnetic resonance (NMR) is a powerful instrumental technique suited to characterize and identify organic substances, and has been successfully applied in the analysis of complex matrices such as biological and environmental samples. In a previous work, we demonstrated the ability of unsupervised contribution analysis (UCA) to process complex mixtures to identify the number of independent constituents and deconvolute mixed signals into specific signal sources. In this work, we evaluated the deconvolving ability of this algorithm to access underlying spectral information—we used UCA to estimate the number of contributing species and respective contributing sources and scores and with that information performed selective 1H-NMR signal suppression. We found that, in optimal NMR conditions, independently of signal source type, UCA allows us to correctly (a) estimate the number of independent contributions, (b) retrieve specific signal sources and (c) respective mixing information, allowing us to (d) characterize each contribution using signal sources and (e) quantify each specific contribution by means of its mixing information. This unsupervised soft-modeling method allows (f) individual contribution estimation and (g) respective removal from collected spectra, thus (h) enhancing spectra information for minor contributing species.