Postharvest performance of apple phenotypes predicted by near-infrared (NIR) spectral analysis

• Post-storage dry-matter (DMC) and soluble solids (SSC) were highly correlated.
• Post-storage DMC and SSC could be predicted well using harvest-time NIR spectra.
• LEGO (leave-each-group/seedling-out in turn) approach was used for validation.
• Efficiency improved by not harvesting, storing and evaluating unwanted seedlings.

To enhance the efficiency of cultivar breeding, the main objective of this study was to develop prediction models using harvest-time spectra to segregate seedlings for postharvest fruit phenotypes. Fruit from 279 genotypes were collected over three seasons, and NIR spectra were recorded using a Nirvana handheld instrument before placing fruit in cool-stores at 0.5 °C for a period of 6–10 weeks. Post-storage spectra were also recorded on all fruit before assessing soluble solids concentration (SSC), dry-matter (DMC), and fruit firmness using destructive techniques. Prediction models were developed using LEGO (leave-each-group-out) approach by leaving out in turn each seedling, and then applying the resulting model to predict postharvest performance of the left-out seedling. DMC was on average about 2% higher than SSC, and a high genetic correlation (0.90) was observed between these two traits. The seedling-level root mean square error of prediction for SSC and DMC were about 0.80% and 0.70% respectively. The correlation between the observed and predicted postharvest performance was 0.91 and 0.95 for SSC and DMC respectively. Estimated heritability and correlation with sensory traits were very similar between the observed and NIR-predicted trait values. Models developed in this study will be used to screen out undesirable seedlings, hence improving efficiency by not harvesting, cool-storing and/or undertaking postharvest sensory evaluation of unwanted seedlings.