Towards a better understanding of the evolution of odour-active compounds and the aroma perception of sparkling wines during ageing

•The biggest differences in most chemical families are related to ageing.•Differential and descriptive tests show differences linked to yeast strain.•PCA models from the OS and SA datasets reveal clusters by ageing and yeast strain.•Floral and fatty attributes respectively achieve the best and worst PLS model.•38 key markers are considered the most influential on sensory attributes. A native veil-forming yeast and a commercial yeast strain were used to elaborate sparkling wines by the Champenoise method with a grape variety traditionally used for the production of still wines. Wines aged on lees for fifteen months were sampled at five points and their physicochemical and sensory indices were analysed. Unsupervised and supervised statistical techniques were used to establish a comparison between 81 volatile compounds and eight odour descriptors (chemical, fruity, floral, fatty, balsamic, vegetal, empyreumatic and spicy). Principal component analysis of both datasets showed good separation among the samples in relation to ageing time and yeast strain. By using a partial least squares regression-based criterion, 38 odour active compounds were selected as the most influential for the ageing factor and out of them, only 27 were unique to certain aroma descriptors. These results contribute to a better understanding of the aroma perception of sparkling wines.