Development of NIR spectroscopy based prediction models for nutritional profiling of pearl millet (Pennisetum glaucum (L.)) R.Br: A chemometrics approach

Abstract

Pearl millet can be viably used for food diversification due to its balanced nutritional composition. Nutritional parameters are conventionally assessed using labour and time-intensive strenuous conventional methods for germplasm screening. Near-infrared reflectance spectroscopy (NIRS) uses near-infrared sections of the electromagnetic spectrum for precise and speedy determination of biochemical parameters for large germplasm. MPLS (Modified Partial Least Squares) regression based NIRS prediction models were developed to assess starch, resistant starch, amylose, protein, oil, total dietary fibre, phenolics, total soluble sugars, phytic acid for high throughput screening of pearl millet germplasm. Mathematical treatments executed by permutation and combinations for calibrating the model, where 2nd, 3rd, and 4th derivatives produced the best results. Treatments “4,5,4,1” was finalized for protein, oil, resistant starch, total dietary fibre, “3,4,4,1” for phenolics, “2,8,4,1” for amylose, “2,4,4,1” for phytic acid, “4,7,4,1” for total soluble sugars and “2,8,4,1” for starch. Treatments with the highest 1-Variance ratio, RSQinternal (coefficient of determination) values, lowest SEC(V) (standard error of cross-validation), SEP(C) (standard error of performance) were identified for subsequent validation. External validation determined the prediction accuracy based on RSQexternal, RPD (residual prediction deviation), SD (standard deviation), p-value ≥ 0.05 and low SEP(C).