Nuclear magnetic resonance methods for the analysis of complex mixtures

Nowadays, a rapidly expanding World drives the development and elaboration of new techniques and methods for fast and efficient analysis. Also, NMR spectroscopy, adapting to the market, wants to keep up with the changes and offers modern experimental solutions. One of today’s needs is the profiling of mixtures. NMR spectroscopy allows their fast and efficient analysis without needing the physical separation of components. However, measurement has to take into account many issues and fine-tune the details. I propose to divide mixtures according to the problems encountered during their analysis. The mixtures portrayed in my dissertation have common denominators: • Mixtures with complicated composition due to the presence of conformers and isotopomers; • Mixtures of a biological nature, with components that hinder analysis because of high peaks overlap; • Mixtures with a dynamically changing composition, whose spectra change due to a chemical reaction carried out directly in the NMR tube or due to the titration-based ligand binding studies; • Mixtures whose spectra are crowded and additional techniques, such as deconvolution, had to be used to aid analysis; • Mixtures whose spectra are affected by temperature or pH change. On the one hand, this enabled the development of additional techniques to support their assignment. On the other hand, for some, this had to be taken into account when designing measurements; • Mixtures with a low concentration of some components whose spectra require sensitivity enhancement. The core of my work consists of six articles published in peer-reviewed journals. They contain a detailed account of the obtained results, conclusions, perspectives, and a thorough comparison with the experiment: 1. Monitoring a reaction causing the resonance frequencies to vary in time. In traditional measurements, this variation causes resolution and sensitivity loss and hampers the detection of transient products’ peaks. I introduced an original approach to correct these non-stationary 2D NMR signals and raise the detection limits. I demonstrate the success of its application in studying the complex reaction mechanism of a radical nature. 2. Usage of NMR experiments combined with computational methods to identify key structural features of S-adenosylmethionine in water. In the dissertation, I discuss the problems associated with NMR-based analysis in deeper detail. The analyzed mixture of conformers provided complex spectra. To analyze their efficiently and determine molecular structure, several measurement conditions had to be optimized. 3. The use of a novel method to overcome the time-consuming NMR-based protein investigation with titration that is non-stationary complementary non-uniform sampling (NOSCO NUS) combined with a robust particle swarm optimization algorithm. We showed its potential in two challenging studies of proteins with different sizes and binding affinities. NOSCO NUS can reduce measurement times and make NMR titration studies more widely usable. 4. 1H NMR-based temperature measurements of metabolomic mixtures and obtaining the temperature coefficients (TCs) of the ingredients (small molecules). I proved that these TCs are characteristic and reproducible between various biological mixtures. I obtained TCs by traditional peak-picking of all spectra in the series and, alternatively, using the Radon transform. The results were comparable, but the Radon transform opened the way to a sensitivity boost in more demanding cases. 5. The proposition of an extension of the 2D Radon spectra analysis (from the previous point) by a U-Net neural network-based peak-picker with a user-friendly graphical interface. We show the software tests on three demanding variable-temperature data sets. Our software deals with peak overlap, low signal-to-noise ratios, and complex multiplet structures. I was responsible for the testing of the program and conceptual design of its interface based on my experience with Radon based mixture analysis. 6. A combination of 1D and 2D single- and triple-quantum techniques to assign all signals in post reaction mixture (unreacted substrates, main products, and side products). This assignment may help analyze the mixtures containing numerous ammonia borane derivatives, which often give overlapping signals that are hard to distinguish. The results published in all six articles forming the core of my dissertation raise the possibility of further development of the proposed methods. Therefore, many prospects for their future use arise.