RUI: Accelerating Advanced Sampling in Multi-Dimensional NMR

This project focuses on the development of advanced data sampling techniques to meet evolving needs in molecular characterization, and focusing on the widely used technique nuclear magnetic resonance (NMR). As NMR contributes more deeply in pharmaceutical sciences, materials and battery research, and even emerging diagnostics, this work will pursue strategies to help scientists throughout these fields deploy new sampling methods without requiring specialization. At the same time, the practice of NMR is being impacted by new machine learning (ML) and deep learning (DL) methods, and this project will investigate how these ML and DL methods interact with modern data sampling strategies in NMR. The research program involves students in valuable training in data science topics while advancing NMR data sampling methods. Alternative data sampling and processing techniques in nuclear magnetic resonance (NMR) yield time and cost savings in multidimensional NMR, while also enhancing the valuable information contained in the data such as through improved resolution. A general approach to such advanced data sampling is termed non-uniform sampling (NUS). Yet there is a need for comparative studies in NUS to guide the design of modern NUS strategies, while metrics are also needed to objectively report on the quality of NUS strategies before investing time at the spectrometer. Metrics to evaluate both 1D-NUS (e.g. of 2D NMR) and nD-NUS (e.g. for 3D-NMR) are needed. This work aims to address such priorities with methods that ultimately aim to help take specialization out of effective NUS practice. This project will also work to build familiarity in the broader community with strengths and weaknesses of NUS in analogy to well-known properties of Fourier transform methods. In light of emerging AI driven methods, fundamental questions on how non-uniform sampling principles (sparsity, decoherence, point spread functions, etc.) relate to the performance of these new methods will be investigated. This project utilizes computation and simulation but also retains a core value around theory-guided solutions. When theory guides new methods, they are then able to perform well prospectively, even in situations where they have not been explicitly tested. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.