Modelling and classification of apple textural attributes using sensory, instrumental and compositional analyses

The textural characteristics of apples impact their quality, storability, and consumer acceptance. While texture can be evaluated instrumentally or sensorially, instrumental measurements are preferred if they can be reliably related to human perception. Therefore, the objectives of this research were to validate instrumental measurements with sensory determinations, develop a classification scheme to group apples by their textural characteristics, and create models to predict sensory attributes from instrumental and compositional analyses. Sensory characteristics (crispness, hardness, juiciness, skin toughness) of 12 apple cultivars were evaluated by 12 judges, in triplicate, on new and established cultivars. Fruit was also evaluated using five instrumental measurements from the TA.XTplus Texture Analyzer and three compositional determinations, on nine unpeeled fruit from each cultivar. The experiment was repeated, to generate two datasets, for analysis and validation processes. T-tests revealed no significant difference between the two datasets, for each of the sensory, instrumental and compositional variables. Principal component analysis (PCA), using the instrumental measurements, revealed that 95.88% of the variation in the instrumental textural determinations could be explained by two components (PC 1, PC 2); these components were highly correlated with flesh firmness and skin strength, respectively. PCA, in combination with the mean sensory scores, revealed four textural groups for the apple cultivars, each with unique sensory characteristics. Four textural groups of apples were identified, and the accuracy of classification was established at 94.44% by using linear discriminant analysis. Multiple regression models were developed and validated to predict sensory textural characteristics from instrumental textural determinations. These predictive models explained more than 85% of variation in the data, for hardness and crispness; while models for juiciness and skin toughness were more complex and explained less variation. This research will assist fruit researchers, apple breeders and industry stakeholders in understanding and predicting the perceived texture of unpeeled apples, using only simple instrumental and basic compositional determinations. The work will allow industry personnel to reduce or eliminate the time-consuming and costly sensory testing – yet have an appreciation of the textural traits as perceived by the consumer.