Department of Food Science and Technology, The Ohio State University, Columbus, OH 43210, USA.
Chempacker LLC, San Jose, CA, USA.
Food Chem. 2023 Mar 30;405(Pt B):134868. doi: 10.1016/j.foodchem.2022.134868. Epub 2022 Nov 7.
In NMR-based untargeted analysis, Fourier transformation is applied to the time-domain data to extract observables such as frequency and intensity. Despite its wide application, this approach has several limitations that can prevent NMR from reaching its highest potential. Here, we utilized Bayesian analysis through CRAFT as an alternative method, using California-style table olives as a model system. Our hypothesis was that the time-domain analysis through CRAFT will be as successful as the traditional approach. The results showed that CRAFT generated efficient unsupervised and supervised models in a robust, and rapid/automated manner. The duration of CRAFT analysis can be further reduced by using the first 14 k complex data points of the initial part of the FID, without affecting the performance of the untargeted analysis. For unsupervised analysis, CRAFT was generally more efficient, while for supervised analysis both approaches were effective. CRAFT can be also used for identifying marker compounds driving classifications.
在基于 NMR 的非靶向分析中，傅立叶变换应用于时域数据以提取可观测值，如频率和强度。尽管该方法应用广泛，但存在一些限制因素，可能会阻碍 NMR 技术发挥其最大潜力。在这里，我们利用贝叶斯分析（通过 CRAFT）作为替代方法，以加利福尼亚式橄榄作为模型系统。我们的假设是，通过 CRAFT 进行的时域分析将与传统方法一样成功。结果表明，CRAFT 以稳健、快速/自动化的方式生成了有效的无监督和有监督模型。通过使用 FID 初始部分的前 14k 个复数数据点，可以进一步缩短 CRAFT 分析的时间，而不会影响非靶向分析的性能。对于无监督分析，CRAFT 通常更有效，而对于有监督分析，两种方法都有效。CRAFT 还可用于识别驱动分类的标记化合物。
