Cysteine Quantification in Pea Cultivars from SERS Spectra Using AI

Rapid quantification of sulfur-containing amino acids, particularly cysteine, in legumes is critical for assessing nutritional quality, supporting breeding program screening, and ensuring consistency in quality control processes. However, conventional methods, such as high-performance liquid chromatography (HPLC), are time-consuming and resource-intensive for high-throughput applications. This study evaluated artificial intelligence models for predicting cultivar-level mean cysteine concentration from surface-enhanced Raman spectroscopy (SERS) spectra of pea extracts. SERS spectra were acquired from 20 cultivars grown at three field locations, with HPLC measurements across the three locations averaged to provide a cultivar-level mean cysteine reference value. Linear regression, partial least squares regression, support vector regression, random forest regression, and a one-dimensional convolutional neural network (1D-CNN) were compared using within-cultivar splits and leave-one-cultivar-out (LOCO) evaluation. The 1D-CNN achieved good predictive performance for the represented cultivars, with an RMSE of 0.005 g/100 g and an MAE of 0.004 g/100 g. Under LOCO evaluation, RMSE and MAE increased to 0.013 and 0.012 g/100 g, respectively. These results provide proof of concept for SERS/AI prediction of multi-environment, cultivar-level mean cysteine concentration, with LOCO providing encouraging evidence of transferability to unseen cultivars within the represented environments and analytical conditions. Shapley Additive Explanations analysis identified contributions from both substrate-related and molecular spectral features. Broader validation using independent samples, locations, extractions, substrates, and reference measurements is required to establish general analytical applicability. A simulated additive-noise sensitivity analysis further evaluated model performance under increasing spectral noise.