Main content area

Amino Acid Quantification in Bulk Soybeans by Transmission Raman Spectroscopy

Schulmerich, Matthew V., Gelber, Matthew K., Azam, Hossain M., Harrison, Sandra K., McKinney, John, Thompson, Dennis, Owen, Bridget, Kull, Linda S., Bhargava, Rohit
Analytical chemistry 2013 v.85 no.23 pp. 11376-11381
Raman spectroscopy, cystine, histidine, least squares, methionine, models, normal values, oils, prediction, protein content, protein sources, protein value, serine, soybeans, tryptophan
Soybeans are a commodity crop of significant economic and nutritional interest. As an important source of protein, buyers of soybeans are interested in not only the total protein content but also in the specific amino acids that comprise the total protein content. Raman spectroscopy has the chemical specificity to measure the twenty common amino acids as pure substances. An unsolved challenge, however, is to quantify varying levels of amino acids mixed together and bound in soybeans at relatively low concentrations. Here we report the use of transmission Raman spectroscopy as a secondary analytical approach to nondestructively measure specific amino acids in intact soybeans. With the employment of a transmission-based Raman instrument, built specifically for nondestructive measurements from bulk soybeans, spectra were collected from twenty-four samples to develop a calibration model using a partial least-squares approach with a random-subset cross validation. The calibration model was validated on an independent set of twenty-five samples for oil, protein, and amino acid predictions. After Raman measurements, the samples were reduced to a fine powder and conventional wet chemistry methods were used for quantifying reference values of protein, oil, and 18 amino acids. We found that the greater the concentrations (% by weight component of interest), the better the calibration model and prediction capabilities. Of the 18 amino acids analyzed, 13 had R² values greater than 0.75 with a standard error of prediction c.a. 3–4% by weight. Serine, histidine, cystine, tryptophan, and methionine showed poor predictions (R² < 0.75), which were likely a result of the small sampling range and the low concentration of these components. It is clear from the correlation plots and root-mean-square error of prediction that Raman spectroscopy has sufficient chemical contrast to nondestructively quantify protein, oil, and specific amino acids in intact soybeans.