Selecting Apples with Better Texture more Quickly
Bejaei, M., Cliff, M.A., & Stanich, K. (2019). Selecting apples with better texture more quickly. Oral and poster sessions presented at BC Tree Fruit Horticultural Symposium, Kelowna, BC.
Plain language summary
Sensory attributes of apple texture are key factors affecting the fruit quality perception and consumers’ acceptance. Researchers in apple breeding programs screen sensory characteristics of thousands of apple selections annually to select cultivars with superior qualities. It is an extremely labour intensive practice, and limited number of fruits are available at earlier stages of selection. The development of predictive models to estimate textural attributes would assist in the screening of new apple selections with desirable traits and maximize the use of resources.
Two experiments were conducted to develop cross-validated sensory models to predict textural attributes of apple using instrumental measurements.
Semi-trained sensory panelists (n = 10) evaluated four textural attributes (crispness, hardness, juiciness, skin toughness) in two experiments using 100-unit continuous linescales in individual booths.
Experiment 1: The Mohr instrument related well to human perception, of the instruments evaluated, and was the instrument of choice for quality assessment of apples. Models for hardness and crispness accounted for a high proportion of the variance (70-82%), with small prediction standard errors (PSEs) (2.67-6.36) (scale 100).
Experiment 2: The hardness model explained more than 90% of the variation using only two variables (Fs, Wf). This indicated that the hardness attribute can easily be used as a screening tool in apple breeding programs for the selection of new apple cultivars.
Results of the principal component analysis using five measurements from TA.XTPlus indicated that 96.8% variation in data could be explained using only two principal components. Apple flesh strength related variables were heavily loaded on the first component while skin strength related variables were heavily loaded on the second component. The findings were verified by correlating the mean textural values with the principal component loadings, such as that performed in internal preference mapping.
The models for hardness and crispness were simple (1- and 2-variables) and accounted for a high proportion of variance (R2 = 85-91%), while those for juiciness and skin toughness were more complex. Cross-validations of the hardness and crispness models were successful (R2 = 79-89%) and provided textural estimates with small standard errors. This meant that the models were practical/appropriate for identifying fruit with superior textural attributes as part of an apple breeding program.