Novel spectroscopic approaches for the characterisation of quality- and identity-related key features of peanuts and peanut butters

Peanuts as raw materials are the cornerstone of the whole peanut industry. The quality traits of peanuts determine their market value and the characteristics of final products, such as peanut butters. However, the quality traits of peanuts vary extensively with variety, growing environment, storage condition, and maturity. Meanwhile, the quality traits of peanuts have profound influences on the characteristics of their derived peanut butters. In order to understand the impact of the varieties on the traits of peanuts and peanut butters and develop rapid determinations, the main aim of this thesis is to elucidate and comprehend distinct analytical signatures and relationships of various types of peanuts and derived peanut butters and develop rapid methods of evaluation and identification of peanuts by near-infrared spectroscopy (NIRS) combined with chemometrics and machine learning.

In Chapter 2 and Chapter 3, the different analytical signatures of batch samples and single peanut kernel peanuts were measured by conventional methods (e.g. gas chromatography and high-performance anion-exchange chromatography). The results showed that the fatty acids (FAs) composition of high oleic acid peanuts (HOP) and regular peanuts differed significantly. The models established by portable NIRS had the same performance to identify HOP and quantitatively measure the major FAs in batch peanut samples compared with benchtop NIRS. Meanwhile, considering the internal (measured by laser confocal microscopy) and external characteristics (kernel size) of different single peanut varieties, a single peanut detection accessory was designed and equipped with portable NIRS to collect spectral data of single peanut kernels. Based on the established quantitative models, breeding experts could quickly analyse the fat, protein, sucrose, and amino acids contents in single peanut kernels.

In Chapter 4, the relationships between peanuts and derived peanut butters were elucidated and the varieties were systematically clustered. It was found that lower fat and higher sucrose content in peanuts had great positive contributions to texture, rheology, and pyrazine content of peanut butters. Amino acids, especially serine, had positive effects on the main volatile compounds. Therefore, one group of peanuts (e.g. HY25, JH18, YH37, etc.) were applied to manufacture peanut butters with the highest pyrazine content and the highest values of texture and rheological properties. In Chapter 5, it was presented that peanut varieties were further clustered according to the structure characteristics (texture and rheology) and roast characteristics (colour and volatile compounds), respectively. The rapid identification methods of peanut varieties for peanut butter manufacture were established based on benchtop NIRS combined with machine learning. The sensitivity, specificity, and accuracy of cross validation and external validations using random forest and support vector machine algorithms were all over 90%, which offered new strategies for producers to rapidly identify peanut varieties for processing purposes.

In conclusion, the obstacles regarding the analytical fingerprints in batch and single kernel measurements as well as the relationship between peanuts and derived peanut butters, were addressed in this thesis. Furthermore, it is expected that the future development of novel spectroscopic approaches based on NIRS could boost critical links to improve the efficiency of peanut quality assessment. From breeding experts to producers, all stakeholders could have deep insights into the impacts of varieties on the quality traits of peanuts and their products. Ultimately, it will provide better control over the quality traits of peanut varieties in order to enhance quality, reduce waste, and mitigate integrity issues in the peanut (butter) supply chains.