Rapid assessment of flavonoid and phenolic contents of tender jackfruit components using miniature near-infrared reflectance spectroscopy: sample presentation and spectral variable selection

This study presented a novel attempt to assess the component-wise composition of total flavonoid content (TFC) and total phenolic content (TPC) in tender jackfruit. It also explored the potential of using miniature near-infrared reflectance spectroscopy (NIRS, 910–1690 nm) for this purpose, with a particular focus on sample presentation (Fresh and Dry samples) and spectral variable selection via the ordered predictor selection (OPS) approach. Results revealed that the mean TPC was significantly higher (α < 0.05) in the tendrils compared to the skin and core components of tender jackfruit, while TFC levels were similar across all components. Partial least squares regression (PLSR) was employed to establish spectra-attribute linkages, and performance was evaluated using the coefficient of determination (R²), root mean square error (RMSE), and residual prediction deviation (RPD) of cross-validation. The spectra from Dry samples proved to be significantly more effective than those from Fresh samples for assessing both TFC (R² = 0.70–0.91; RMSE = 2.42–4.72 mg RE/g) and TPC (R² = 0.81–0.92; RMSE = 2.24–3.68 mg GAE/g) across different tender jackfruit components. The OPS-selected variables outperformed the conventional full-spectrum PLSR models for both TFC and TPC irrespective of components. The OPS approach resulted in excellent models (RPD > 3) for both phytochemicals in the skin and tendrils, and models of good accuracy (2.5 < RPD < 3) for the core components. Based on these findings, the use of miniature NIRS, combined with Dry samples and the OPS approach, may be recommended as a promising tool for rapid and reliable assessment of TFC and TPC in tender jackfruit components.