Classification of essential oil composition in Rosa damascena Mill. genotypes using an electronic nose

Identificador:  imarina:9285496

Handle: https://hdl.handle.net/20.500.11797/imarina9285496

Autors:  Gorji-Chakespari A; Nikbakht AM; Sefidkon F; Ghasemi-Varnamkhasti M; Valero EL

Resum:
One of the major problems in the industry of medicinal and aromatic plants (MAPs) is the absence of a quick, easy and inexpensive method for controlling the quality of these plants. Rosa damascena Mill., is an aromatic plant which is cultivated for its high-value essential oil in the world. In this study, essential oils were extracted from nine genotypes of Rosa, and their main components were identified by GC and GC–MS. Then, the samples from different genotypes were grouped in three classes (C1, C2, C3) based on their total percentage of the six most important constituents, which have a major role in the quality of essential oil (i.e., phenyl ethyl alcohol, trans rose oxide, citronellol, nerol, geraniol, geranial). An electronic nose (EN) system was designed based on metal oxide semiconductor (MOS) sensors, and trained to identify the categories to which samples of essential oils could be classified. The response patterns of the sensors were recorded and further analyzed by chemometrics methods. Based on the results, principal components analysis (PCA) and linear discrimination analysis (LDA) showed that 85% and 99% of sample variance could be explained by the first two principal components (PC1, PC2) and two linear discrimination axis (LD1, LD2), respectively. LDA was performed on sensor response variables by cross- validated dataset (5- fold) and the classification accuracy was 95%. Finally, an error-correcting output codes (ECOC) classifier as a multiclass model for support vector machines (SVM) was considered and the classification accuracy was increased to 99%. These results reveal that an EN can be used as a quick, easy, accurate and inexpensive system for the classification of essential oil composition in Rosa damascena Mill. © 2016 Elsevier GmbH