Potential of spectroscopic analyses for non-destructive estimation of tea quality-related metabolites in fresh new leaves

www.nature.com/scientificreports OPEN Potential of spectroscopic analyses for non‑destructive estimation of tea quality‑related metabolites in fresh new leaves Hiroto Yamashita1,2, Rei Sonobe1,3*, Yuhei Hirono3,4, Akio Morita1,3 & Takashi Ikka1,3* Spectroscopic sensing provides physical and chemical information in a non-destructive and rapid manner. To develop non-destructive estimation methods of tea quality-related metabolites in fresh leaves, we estimated the contents of free amino acids, catechins, and caffeine in fresh tea leaves using visible to short-wave infrared hyperspectral reflectance data and machine learning algorithms. We acquired these data from approximately 200 new leaves with various status and then constructed the regression model in the combination of six spectral patterns with pre-processing and five algorithms. In most phenotypes, the combination of de-trending pre-processing and Cubist algorithms was robustly selected as the best combination in each round over 100 repetitions that were evaluated based on the ratio of performance to deviation (RPD) values. The mean RPD values were ranged from 1.1 to 2.7 and most of them were above the acceptable or accurate threshold (RPD = 1.4 or 2.0, respectively). Data-based sensitivity analysis identified the important hyperspectral regions around 1500 and 2000 nm. Present spectroscopic approaches indicate that most tea quality-related metabolites can be estimated non-destructively, and pre-processing techniques help to improve its accuracy. Plants collectively produce many metabolites with estimates ranging from 100,000 to 1 million, and many metabolites are thought to play essential roles in resistance to biotic stresses and tolerance of abiotic stresses1–5. In addition, natural products synthesized in plants provide indispensable resources for human health and survival5. Given the importance of plant metabolites to plant development and adaptation, and for human health, various quantitative and qualitative analyses have been developed. The main examples are based on chromatography techniques such as gas chromatography or high-performance liquid chromatography (HPLC) with improved mass resolution and s ensitivity6,7. However, these analytical methods require the destructive collection and pretreatment of plant samples, which makes them slow in acquiring analytical data and unsuitable for real-time diagnosis of metabolite level. Hyperspectral reflectance sensing is an established spectroscopic method that can provide rapid analysis without the need for sample pre-treatment. It is commonly applied to visible (VIS; 400–700 nm), near-infrared (NIR; 700–1000 nm), and short-wave infrared (SWIR; 1000–2500 nm) spectral ranges and has been used to estimate leaf pigments and water contents8,9. The VIS is dominated by absorption of the photosynthetic pigments such as chlorophylls, carotenoids, and a nthocyanins8. On the other hand, NIR spectroscopy is directly relevant to the overtones and combinations of the fundamental C–H, O–H, and N–H bonds in organic m olecules10,11. Thus, NIR spectroscopy provides physical and chemical information and has shown good potential in estimating different parameters in biotic samples, including metabolites in plants, agricultural products, and f ood12–14. In addition, machine learning techniques provide powerful tools for constructing regression or classification models in agricultural indices from hyperspectral reflectance d ata15. The methodology of machine learning algorithms provides a flexible model not only for data-driven decision-making but also for capturing expertise into the algorithms16. The technique shows good potential for analyzing hyperspectral reflectance data with all 1 Faculty of Agriculture, Shizuoka University, 836 Ohya, Suruga‑ku, Shizuoka 422‑8529, Japan. 2United Graduate School of Agricultural Science, Gifu University, 1‑1 Yanagito, Gifu 501‑1193, Japan. 3Institute for Tea Science, Shizuoka University, 836 Ohya, Suruga‑ku, Shizuoka 422‑8529, Japan. 4Division of Tea Research, Institute of Fruit Tree and Tea Science, National Agriculture and Food Research Organization (NARO), 2769 Shishidoi, Kanaya, Shimada, Shizuoka 428‑8501, Japan. *email: ; Scientific Reports | (2021) 11:4169 | https://doi.org/10.1038/s41598-021-83847-0 1 Vol.:(0123456789) www.nature.com/scientificreports/ spectral information based on a large number of bands17. Machine learning techniques also enable the assessment of hyperspectral features that are informative for high accuracy predictive m odelling16,18. Tea plants (Camellia sinensis L.) are mainly distributed and cultivated in Asia to produce several tea types, such as green tea, oolong tea, and black tea, which are popular non-alcoholic beverages consumed all over the world. Tea-drinking reportedly has numerous and diverse health benefits19. Generally, tea quality and function are defined by the profile of various chemical components, such as catechins, caffeine, and theanine, which are characteristics to tea leaves. Tea catechins, which comprise a major class of polyphenols, contribute to the taste of astringency and bitterness of tea and have been studied for their health functions such as antibacterial a ctivities20 and free radical scavenging a ctivities21. Free amino acids, especially glutamate (Glu) and theanine, contribute to the umami taste of green tea22,23. In particular, theanine, a unique amino acid in tea plants, has the activities of promoting r elaxation24 and reducing blood p ressure25. Caffeine (1,3,7-trimethylxanthine) is a kind of purine alkaloid and its consumption may be associated with a reduced risk for type 2 diabetes26, but excessive intake of caffeine may cause inflammation of the digestive organs, insomnia, and arrhythmia27. Thus, unique tea qualityrelated metabolites are the most important agronomic traits targeted by modern and future tea cultivation and breeding. To evaluate the levels of these metabolites, many analytical tools have been employed to quantify tea quality-related metabolites including free amino acids, catechins, and caffeine contents in tea samples. Many analytical methods have been based on HPLC28,29 and capillary electrophoresis30,31, but these methods destructively use plant tissues and are time-consuming and expensive to perform. Therefore, a rapid and accurate method for the evaluation of quantitative traits in tea leaves is in high demand for tea cultivation management and breeding programs. The NIR-based estimation of some chemical components in ground tea leaves has been established by previous s tudies32–34. Few studies have been reported in a non-destructive method for fresh l eaves35,36. Huang et al.35 have reported non-destructive estimation methods for four main catechins and caffeine in fresh green leaves based on VIS–NIR spectra (400–2498 nm) and partial least squares (PLS) model. However, the outcomes of this study were limited by fewer tea quality-related meta (...truncated)