Near-Infrared Spectroscopy Prediction of Dry Matter and Starch Content in Cassava Using Optimized Calibration Models.

Area of Science:

• Agricultural Science
• Spectroscopy
• Plant Breeding

Background:

• Dry matter content (DMC) and starch content (StC) are crucial for cassava quality.
• Traditional phenotyping methods for these traits are labor-intensive and hinder breeding program scalability.

Purpose of the Study:

• To develop and compare predictive models for DMC and StC using near-infrared (NIR) spectroscopy.
• To evaluate two NIR devices (benchtop and portable) and the impact of sample type (fresh vs. processed).

Main Methods:

• Analysis of 3,391 cassava clones using spectral data (1000-2500 nm and 350-2500 nm).
• Development of predictive models with Partial Least Squares (PLS), k-Nearest Neighbors (KNN), and eXtreme Gradient Boosting (XGB).
• Comparison of model performance based on sample preparation and device type.

Main Results:

• Partial Least Squares (PLS) models demonstrated high predictive accuracy across traits and devices.
• Processed samples consistently yielded higher model accuracy compared to fresh samples.
• The portable NIR device showed competitive and sometimes superior performance, especially with processed samples in external validation.

Conclusions:

• NIR spectroscopy is a viable tool for rapid and accurate phenotyping of cassava quality traits.
• Processed sample preparation significantly enhances predictive model performance.
• Portable NIR spectrometers offer a practical and scalable alternative for high-throughput cassava breeding.