Multivariate statistical visualisation and modelling of GC–MS data for sensory quality prediction of flavoured cider as influenced by storage time and temperature

Abstract

It is important to cider producers to ensure the optimal quality of their products directly after production, and after extended shelf life. Quality assessments usually involve regular and continuous sensory evaluation conducted by trained panels. The competitive cider production environment necessitates extensive sensory analysis for new product launches, as part of quality control of existing products and during the production process. This could potentially lead to panel fatigue, less accuracy in sensory evaluation and delays in product launches. This paper describes the development of a partial least squares (PLS) model for the prediction of cider quality based on multivariate correlation of sensory and gas chromatography-mass spectroscopy (GC–MS) volatile compound profile data of a flavoured cider. Clear differences between cider samples of various batches and storage conditions could be visualised using principal component analysis (PCA). However, the developed PLS model was able to predict the commercial cider’s quality irrespective of differences in batch, storage temperature or storage time. The main compounds that are strongly associated with sensory quality and its decline during shelf-life storage were identified. This model is the first quality prediction model to be developed for South African flavoured cider which could be used to facilitate fast, consistent and accurate quality assessment in the high-pressure cider production and trade environment. Furthermore, since the model captures the sensory data obtained from a trained technical panel, it can be used to facilitate the calibration of sensory panels across production sites, reduce product development time and better optimise processes that impact sensory quality.