Flavour profiling of fresh and processed fruit smoothies by instrumental and sensory analysis

Flavour volatile compounds and descriptive sensory attributes were examined in fresh, thermal and high hydrostatic pressure (HHP) processed (450 MPa/5 min/20 °C or 600 MPa/10 min/20 °C) fruit smoothies stored over 30 days (4 °C). Volatile compounds were extracted by solid-phase microextraction (SPME) and analysed by gas chromatography mass spectrometry (GCMS). Limonene and trans-2-hexanal were the most abundant compounds present in the smoothies. Processing affected the limonene content of smoothies, with thermally processed samples shown to be higher (p < 0.05) than all other samples. As expected, storage decreased (p < 0.05) the volatile content of all smoothies, in particular for volatiles associated with 'fresh/green/grassy' flavour notes, like trans-2-hexanal. Principal component analysis (PCA) of descriptive sensory data revealed that attributes such as pink colour, fresh colour, apple aroma, fresh aroma, and fresh flavour were more associated with both fresh and processed smoothies on day 0. Physico-chemical assessment showed that colour was affected (p < 0.05) by both processing and storage but the changes were not consistent for the individual processing types i.e. HHP treated samples showed good initial colour retention but had the biggest overall changes over time, while thermal treatments initially degraded colour but had the lowest colour change over time. Partial least squares (PLS) regression was used to investigate the relationships between sensory attributes and volatile compounds. However, correlations between sensory and instrumental data were lower than expectations, possibly due to the high odour threshold values of the compounds identified.

► Processing/storage effects on sensory characteristics of smoothies were evaluated. ► Processing & storage affected volatiles extracted by SPME and analysed by GCMS. ► PCA of sensory data sorted fresh & processed samples (d0) with 'fresh' descriptors. ► Antioxidant contents fitted well to second order polynomial models. ► PLSR of sensory attributes & instrumental volatiles were lower than expected.