Skip to main content
ARS Home » Plains Area » Manhattan, Kansas » Center for Grain and Animal Health Research » Grain Quality and Structure Research » Research » Publications at this Location » Publication #411876
Research Project: Grain Composition Traits Related to End-Use Quality and Value of Sorghum

Location: Grain Quality and Structure Research

Title: Estimating sorghum leaf dhurrin levels using a handheld near infrared instrument

Author
item Peiris, Kamaranga
item Bean, Scott
item HAYES, CHAD - US Department Of Agriculture (USDA)
item EMENDACK, YVES - US Department Of Agriculture (USDA)
item SANCHEZ, JACOBO - US Department Of Agriculture (USDA)

Submitted to: Meeting Abstract
Publication Type: Abstract Only
Publication Acceptance Date: 4/1/2024
Publication Date: N/A
Citation: N/A

Interpretive Summary:

Technical Abstract: Measurement of dhurrin levels in sorghum germplasm is important to assess suitability as a forage crop and for evaluation of pre and post flower drought tolerance. Dhurrin levels are typically measured by high-performance liquid chromatography (HPLC), which is laborious, time consuming, and expensive. To reduce analysis time, we tested the feasibility of using near-infrared spectroscopy for estimating leaf dhurrin levels using a handheld NIR. A sample set of 150 sorghum leaves consisting of three leaves (first leaf below the flag leaf) from 50 varieties were harvested at physiological maturity and leaf disks were collected for dhurrin measurements by HPLC. For NIR analysis, leaves were scanned after folding the leaf two or three times (4 or 8 leaf blades) and 6 scans were taken from each fold using a handheld NIR instrument. After the fresh leaves were scanned, the leaves were left to dry on the benchtop for 10 days. Dried leaves were then scanned in the same way as the fresh leaves Each leaf was considered as an independent sample and dhurrin levels of 100 leaves were used for calibration development and 50 for validation. Calibration models were developed using the spectra of leaves and the respective dhurrin levels of leaves on fresh weight basis. The calibration models built with NIR scans from fresh leaves did not produce useful models probably due to masking of the dhurrin absorption bands by strong moisture bands. Several PLS regression models, back propagation neural network (BPNN) and deep learning neural networks were evaluated for calibration development. BPNN model predicted the dhurrin levels of the validation samples with a RMSEP of 5.95 ug/g. PLS calibration predicted the validation set with RMSEP of 6.10 ug/g. The results of this experiment showed that NIR spectroscopy is feasible for estimation of dhurrin levels in dried sorghum leaves.