Résumé
To be able to account for sensory quality earlier in the assessment of a new banana hybrid in a selection scheme, the ability of instrumental parameters to predict the sensory perception of banana odour and aroma was investigated. Thirteen cultivated bananas and four new triploid hybrids were characterized by both sensory profiling and chemical analyses. PLS regression models were built using cultivated bananas, and the quality of predictions was validated using hybrids. Forty-one volatile compounds were detected in banana cultivars by SPME-GC-MS. Major differences were observed in the distribution of volatile compounds among cultivars. 3-methylbutyl esters were the most abundant in the 17 banana cultivars. PLS models were able to predict four odours ('banana', 'fermented', 'grassy', and 'chemical') and five aromas ('banana', 'fermented', 'chemical', 'medicinal', and 'pineapple') with a goodness of fit (R-2) between 0.52 and 0.71. Two butanoate esters, 2-methylpropyl butanoate and 3-methylbutyl butanoate, mainly contributed to 'banana' odour and aroma. Two other esters, 3-methylbutyl acetate and ethyl 3-methylbutanoate contributed to the 'fermented', 'chemical', and 'medicinal' notes, suggesting that interactions between volatile compounds could lead to differences in the perception of odours and aromas. The 'grassy' odour was the only attribute which was negatively associated with volatile compounds. Total organic acids enhanced the 'pineapple' aroma, whereas it reduced the perception of the 'banana' aroma, suggesting interactions between volatile and non-volatile compounds or cognitive associations between aromas and sourness. In conclusion, PLS models can now be used for high throughput phenotyping to predict odours and aromas in dessert banana. (C) 2015 Elsevier B.V. All rights reserved.