Authenticity of cinnamon verum determined by MALDI TOF MS and machine learning

Cinnamon is the most popular spice globally and has several potential medicinal applications. Cinnamomum verum is the most valuable and called ‘true cinnamon’ and is mainly produced in Sri Lanka. Other cinnamon spices, Cinnamomum cassia are also commercialized as low cost and contain a higher amount of hepatoxic compound coumarin. On the other hand, the higher price of C. verum makes it more vulnerable to adulteration mainly through the addition of products of the same species. This study investigates whether MALDI-TOF-MS combined with machine learning algorithms is an easy, rapid, and cost-effective analytical method for the authentication of cinnamon. We purchased 10 types of cinnamon powder from different sources, among which four were Cinnamomum verum and the remaining 6 were Cinnamomum cassia. Cinnamon samples were extracted using a solvent and spotted on a MALDI plate. Machine learning algorithms have been used to detect patterns embedded within MALDI-TOF spectra using a training set of 384 spots. Machine learning models were used to classify a different scoring set consisting of 328 sample spots. The approach's sensitivity (sens) and specificity (spc) have been evaluated. After analyses of different machine learning models, the random forest model gave the best results, with the maximal Youden index of J = 1 achieved for the given set of conditions. Gradient boosted tree and decision tree models also gave outstanding results with maximal Youden index of J = 0.99 and 0.98 respectively. We conclude that MALDI TOF MS combined with the machine learning model is a valuable tool for the authentication of cinnamon.