Determination of protein and fatty acid composition of shell-intact upland cottonseed using near-infrared reflectance spectroscopy

Abstract

Upland cotton is a leading natural fiber crop; and an important source of vegetable oil and protein. The protein content and fatty acid composition are important nutritional indicators of cottonseed. In order to establish rapid, high throughout, nondestructive and low-cost method for cottonseed nutrition evaluation, phenotypic data of seed kernel protein, oil and fatty acid content were combined with preprocessed near-infrared spectra of shell-intact cottonseed from 179 upland cotton genotypes to establish calibration models for these cottonseed nutrients. A total of 17 kinds of fatty acids were stably detected and quantitatively determined using Gas Chromatography-Mass Spectrometry (GC-MS) technology, including 7 kinds of fatty acids which were detected in cottonseed, to the best of our knowledge, for the first time. Finally, 19 NIR calibration models were established through partial least squares regression analysis, and evaluated by cross-validation and external verification. Results indicated that 16 models for seed protein, oil, and 14 fatty acids (capric acid, lauric acid, myristic acid, pentadecanoic acid, palmitic acid, heptadecanoic acid, linoleic acid, arachidic acid, cis-11-eicosenoic acid, linolenic acid, heneicosanoic acid, docosanoic acid, tridecanoic acid, and lignoceric acid) achieved acceptable or satisfactory prediction ability. Three calibration models for palmitoleic acid, stearic acid and oleic acid obtained moderate prediction ability with R2cal above 0.7 which could be used in rough screening. Nine fatty acids in cottonseed were modeled for the first time. These results broadened our understanding of the fatty acid profile of cottonseed oil and laid a foundation for fast, nondestructive evaluation of cottonseed nutrient quality for cotton breeding, processing and marketing.