Figure 2: NMR measurements and (pseudo) two-dimensional mapping with Clustering NMR approach. T1 and T2relaxation times were carried on various edible oils (i.e., peanut, olive, sunflower, corn), with the label of (A, B, C, D) respectively using micro NMR relaxometry. One-dimensional mapping with (a) T1 relaxation time, (b) T2 relaxation time, and (c) (pseudo) two-dimensional mapping using a pair of (T1, T2) relaxation times. NMR measurements were carried out on each edible oils in quintuplicate manner with a total of 40 points (datasets). The clustering circles were drawn for eye-balling purposes. Details of the oils (e.g., manufacturers, fat compositions) were presented in Supp. Fig. 1. The box plots represent 25% and 75% quantile of the entire measurements. The diaganol line (T1=T2) represents the border limit where it is physically non-measurable. Two tailed Student“s T-test was used to calculate the P -value.
NMR measurements were carried out (in single blinded manner) on each oils in quintuplicate manner (i.e., five repeated times) with a total of 40 points for all the samples. Details on NMR parameter are reported in Supplementary Methods. Clustering NMR method uses a pair of (T2, T1) relaxation time for each objects (e.g., edible oils, blood) to construct a (pseudo) two-dimensional map (Fig. 2c). The pseudo two-dimensional map can be used a referencing map (control).
Machine learning learning algorithm and workflows. Using a statistical programming languages (e.g., R or Orange 3.1.2), the raw datasets can be processed using supervised and unsupervised learning techniques. The machine learning algorithms were written and runs on a personal laptop (Intel Core Pentium i7 CPU @ 2.70GHz, 8.00 GB RAM). Once the model in machine learning is built, all the tasks run simultaneously and completes typically in less than 1 minute.