Dataset Information

Evaluation of Leymus chinensis quality using near-infrared reflectance spectroscopy with three different statistical analyses.

ABSTRACT: Due to a boom in the dairy industry in Northeast China, the hay industry has been developing rapidly. Thus, it is very important to evaluate the hay quality with a rapid and accurate method. In this research, a novel technique that combines near infrared spectroscopy (NIRs) with three different statistical analyses (MLR, PCR and PLS) was used to predict the chemical quality of sheepgrass (Leymus chinensis) in Heilongjiang Province, China including the concentrations of crude protein (CP), acid detergent fiber (ADF), and neutral detergent fiber (NDF). Firstly, the linear partial least squares regression (PLS) was performed on the spectra and the predictions were compared to those with laboratory-based recorded spectra. Then, the MLR evaluation method for CP has a potential to be used for industry requirements, as it needs less sophisticated and cheaper instrumentation using only a few wavelengths. Results show that in terms of CP, ADF and NDF, (i) the prediction accuracy in terms of CP, ADF and NDF using PLS was obviously improved compared to the PCR algorithm, and comparable or even better than results generated using the MLR algorithm; (ii) the predictions were worse compared to laboratory-based spectra with the MLR algorithmin, and poor predictions were obtained (R2, 0.62, RPD, 0.9) using MLR in terms of NDF; (iii) a satisfactory accuracy with R2 and RPD by PLS method of 0.91, 3.2 for CP, 0.89, 3.1 for ADF and 0.88, 3.0 for NDF, respectively, was obtained. Our results highlight the use of the combined NIRs-PLS method could be applied as a valuable technique to rapidly and accurately evaluate the quality of sheepgrass hay.

SUBMITTER: Chen J

PROVIDER: S-EPMC4671155 | biostudies-literature | 2015

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Evaluation of Leymus chinensis quality using near-infrared reflectance spectroscopy with three different statistical analyses.

Chen Jishan J Zhu Ruifen R Xu Ruixuan R Zhang Wenjun W Shen Yue Y Zhang Yingjun Y

PeerJ 20151203

Due to a boom in the dairy industry in Northeast China, the hay industry has been developing rapidly. Thus, it is very important to evaluate the hay quality with a rapid and accurate method. In this research, a novel technique that combines near infrared spectroscopy (NIRs) with three different statistical analyses (MLR, PCR and PLS) was used to predict the chemical quality of sheepgrass (Leymus chinensis) in Heilongjiang Province, China including the concentrations of crude protein (CP), acid d ...[more]

PMID: 26644973

Similar Datasets

Project description:Optimal management of free-ranging herbivores requires the accurate assessment of an animal's nutritional status. For this purpose 'near-infrared reflectance spectroscopy' (NIRS) is very useful, especially when nutritional assessment is done through faecal indicators such as faecal nitrogen (FN). In order to perform an NIRS calibration, the default protocol recommends starting by generating an initial equation based on at least 50-75 samples from the given species. Although this protocol optimises prediction accuracy, it limits the use of NIRS with rare or endangered species where sample sizes are often small. To overcome this limitation we tested a single NIRS equation (i.e., multispecies calibration) to predict FN in herbivores. Firstly, we used five herbivore species with highly contrasting digestive physiologies to build monospecies and multispecies calibrations, namely horse, sheep, Pyrenean chamois, red deer and European rabbit. Secondly, the equation accuracy was evaluated by two procedures using: (1) an external validation with samples from the same species, which were not used in the calibration process; and (2) samples from different ungulate species, specifically Alpine ibex, domestic goat, European mouflon, roe deer and cattle. The multispecies equation was highly accurate in terms of the coefficient of determination for calibration R2 = 0.98, standard error of validation SECV = 0.10, standard error of external validation SEP = 0.12, ratio of performance to deviation RPD = 5.3, and range error of prediction RER = 28.4. The accuracy of the multispecies equation to predict other herbivore species was also satisfactory (R2 > 0.86, SEP < 0.27, RPD > 2.6, and RER > 8.1). Lastly, the agreement between multi- and monospecies calibrations was also confirmed by the Bland-Altman method. In conclusion, our single multispecies equation can be used as a reliable, cost-effective, easy and powerful analytical method to assess FN in a wide range of herbivore species.

Project description:BackgroundGlucobrassicin (GBS) and its hydrolysis product indole-3-carbinol are important nutritional constituents implicated in cancer chemoprevention. Dietary consumption of vegetables sources of GBS, such as cabbage and Brussels sprouts, is linked to tumor suppression, carcinogen excretion, and cancer-risk reduction. High-performance liquid-chromatography (HPLC) is the current standard GBS identification method, and quantification is based on UV-light absorption in comparison to known standards or via mass spectrometry. These analytical techniques require expensive equipment, trained laboratory personnel, hazardous chemicals, and they are labor intensive. A rapid, nondestructive, inexpensive quantification method is needed to accelerate the adoption of GBS-enhancing production systems. Such an analytical method would allow producers to quantify the quality of their products and give plant breeders a high-throughput phenotyping tool to increase the scale of their breeding programs for high GBS-accumulating varieties. Near-infrared reflectance spectroscopy (NIRS) paired with partial least squares regression (PLSR) could be a useful tool to develop such a method.ResultsHere we demonstrate that GBS concentrations of freeze-dried tissue from a wide variety of cabbage and Brussels sprouts can be predicted using partial least squares regression from NIRS data generated from wavelengths between 950 and 1650 nm. Cross-validation models had R2 = 0.75 with RPD = 2.3 for predicting µmol GBS·100 g-1 fresh weight and R2 = 0.80 with RPD = 2.4 for predicting µmol GBS·g-1 dry weight. Inspections of equation loadings suggest the molecular associations used in modeling may be due to first overtones from O-H stretching and/or N-H stretching of amines.ConclusionsA calibration model suitable for screening GBS concentration of freeze-dried leaf tissue using NIRS-generated data paired with PLSR can be created for cabbage and Brussels sprouts. Optimal NIRS wavelength ranges for calibration remain an open question.

Dataset Information

Evaluation of Leymus chinensis quality using near-infrared reflectance spectroscopy with three different statistical analyses.

Publications

Evaluation of Leymus chinensis quality using near-infrared reflectance spectroscopy with three different statistical analyses.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets