Abstract
Roasting intensity and cultivar shape the physicochemical composition and sensory characteristics of date seed-based coffee alternatives. This study evaluated quality traits among eight date seed cultivars (Zahidi, Medjool, Deglet nour, Thoory, Halawi, Barhee, Khadrawy, Bau Strami) roasted at three intensities (light: 180 °C; medium: 200 °C; dark: 220 °C) using digital technologies, including near-infrared spectroscopy (NIR), electronic nose (e-nose), and headspace solid-phase microextraction gas chromatography-mass spectrometry (HS-SPME-GC-MS), supported by machine learning (ML) modeling. NIR spectra showed distinct chemical fingerprints for date seed powders and beverages, with key absorption bands from 1673-2396 nm and 1720-1927/2238-2396 nm, respectively. E-nose outputs showed higher volatile emissions in dark-roasted samples, particularly for ethanol and NH(3). GC-MS identified 25 volatile compounds, mainly pyrazines and furanic compounds. Pyrazine concentration was greatest in Bau Strami and Medjool cultivars, whereas Halawi and Thoory cultivars had greater content of furfural. Two ML classification models achieved high accuracy in classifying cultivars (NIR inputs: 99%; e-nose inputs: 98%) and roasting levels, while regression models (NIR inputs: R = 0.88; e-nose inputs: R = 0.90) effectively predicted volatile aromatic compounds obtained using GC-MS. Dark roasting resulted in a significant pH reduction and intensified browning, with furfural persisting as a stable aroma contributor. These findings highlight the potential of date seeds as a coffee alternative, with roasting level and cultivar selection influencing flavor profiles. The findings also demonstrate the utility of digital sensing technologies as an efficient, low-cost tool for rapid quality assessment and process optimization in the development of novel beverages.