Development of NIRS models to predict composition of enzymatically processed sweetpotato

This study was conducted to develop calibration models to predict the major constituents (moisture, protein, fiber, alcohol insoluble solids (AIS), and starch) of enzymatically processed sweetpotatoes using a non-destructive near-infrared spectroscopy (NIRS) technique. Prediction of these constituents is of interest since starch content can be used to estimate crop potential and efficiency of processing enzymes used to convert starch into valuable products needed for industrial applications. Wet chemistry procedures are expensive, laborious, and time consuming; however, NIRS is a reliable and fast tool that can be used to quantify components and identify composition changes occurring during sweetpotato processing. Freeze-dried samples of sweetpotato roots (clones: NC-413, DM02-180, and Covington) were scanned over the near infrared wavelengths at different stages of processing (unprocessed material, wet samples after liquefaction, and wet samples after saccharification) and chemically analyzed. Calibration models were established by Multiple Linear Regression (MLR) analysis and developed to predict moisture, AIS, protein, fiber, and starch content. Spectral range and the number of MLR factors were examined in a stepwise manner that yielded the lowest standard error of calibration (SEC) and highest correlation coefficient of determination (R2). Calibration models based on all sweetpotato clones adequately predicted moisture, AIS, and starch compounds in unprocessed and processed treatments. Protein was successfully predicted with 99% confidence for unprocessed material and an approximate quantitative prediction in processed treatments (R2 = 0.69). Fiber was predicted with 85% confidence for Covington sweetpotato and with 65% for both NC-413 and DM02-180 sweetpotato clones. Starch was successfully predicted with 91% and 97% confidence for unprocessed and processed treatments, respectively. Our results indicated that NIRS technique is a tool able to rapidly predict with reasonable accuracy the composition of different constituents present in sweetpotato samples before and during its processing to value-added products.