Non‐Uniform Sampling for Quantitative NOESY

ABSTRACT Non‐uniform sampling (NUS) enables faster acquisition of NMR spectra. Concerns about spectral fidelity, particularly in high‐dynamic‐range experiments like NOESY, have limited its quantitative applications. In this study, we assessed whether optimised Poisson‐gap sampling schemes can generate high‐fidelity spectra suitable for quantitation and evaluated the effectiveness of NUS ranking tools, NUSscore and nus‐tool, in identifying optimal sampling schemes. A total of 25,000 Poisson‐gap sampling schemes were generated and ranked using NUSscore, with a subset of 11 of these spanning the score distribution, alongside 15 random‐shuffle and the highest and lowest scoring Poisson‐gap schemes determined using the signal apex‐to‐artefact ratio were used for comparison, all with 50% sampling coverage. Additionally, hybrid sampling schemes incorporating a long initial uniformly sampled section, termed US‐NUS hybrid schemes, were evaluated. Spectral fidelity was evaluated on interproton distance accuracy, including the proportion of retained interproton distances and their deviation from uniformly sampled reference spectra. NUSscore showed a strong correlation with spectral fidelity. The peak‐to‐sidelobe ratio implemented in nus‐tool showed no correlation, with the relative sensitivity metric showing a weak correlation. Signal‐to‐artefact apex ratio was also not predictive for identifying sampling schedules with maintained interproton distances. All Poisson‐gap sampling schemes however outperformed random‐shuffle. The US‐NUS hybrids demonstrated improved interproton distance conservation than traditional Poisson‐gap sampling schemes with a low seed dependence, making them a promising sampling schedule for quantitative NOESY analysis.