Rapid monitoring of drought and salinity stress responses in wheat via potential Raman-derived biomarkers and traditional biochemical indicators.

Abiotic stresses such as drought and salinity significantly constrain the productivity of in vitro-grown wheat (Triticum aestivum L.) by disrupting its biochemical and physiological homeostasis. Rapid, non-destructive, and data-driven diagnostic approaches are therefore essential for the early detection of stress conditions and for supporting sustainable crop management. In this study, Raman spectroscopy (RS) was integrated with conventional biochemical assays to investigate wheat responses under controlled drought and salinity stress treatments. Distinct Raman spectral features associated with pigments, proteins, carbohydrates, and lipids were analyzed alongside biochemical indicators, including proline, chlorophyll, and malondialdehyde levels. Overall, the integration of RS with machine learning provides a rapid, robust, and non-invasive framework for the early detection of drought and salinity stress in wheat. Notably, Raman intensity variations observed at 737, 996, 1051, 1064, and 1518 [Formula: see text] exhibited consistent spectral trends that closely mirrored changes in conventional biochemical stress markers, confirming that these spectral shifts directly reflect underlying physiological stress responses. To classify stress levels and to identify key Raman-derived biomarkers associated with each stress type, a machine learning approach was implemented, achieving a classification accuracy exceeding 85% in discriminating control, drought-stressed, and salinity-stressed plants. Furthermore, characteristic Raman bands, particularly those associated with C-H and amide vibrational modes, showed strong correlations with established biochemical indicators, underscoring their potential as reliable, non-invasive stress biomarkers. Collectively, these findings provide mechanistic insight into stress-induced structural and biochemical alterations and support the application of RS-machine learning integration for precision agriculture and resilient crop management under changing environmental conditions.