Machine Learning Analysis of Hyperspectral Images of Damaged Wheat Kernels.

Fusarium head blight (FHB) is a disease of small grains caused by the fungus Fusarium graminearum. In this study, we explored the use of hyperspectral imaging (HSI) to evaluate the damage caused by FHB in wheat kernels. We evaluated the use of HSI for disease classification and correlated the damage with the mycotoxin deoxynivalenol (DON) content. Computational analyses were carried out to determine which machine learning methods had the best accuracy to classify different levels of damage in wheat kernel samples. The classes of samples were based on the DON content obtained from Gas Chromatography-Mass Spectrometry (GC-MS). We found that G-Boost, an ensemble method, showed the best performance with 97% accuracy in classifying wheat kernels into different severity levels. Mask R-CNN, an instance segmentation method, was used to segment the wheat kernels from HSI data. The regions of interest (ROIs) obtained from Mask R-CNN achieved a high mAP of 0.97. The results from Mask R-CNN, when combined with the classification method, were able to correlate HSI data with the DON concentration in small grains with an R2 of 0.75. Our results show the potential of HSI to quantify DON in wheat kernels in commercial settings such as elevators or mills.

Physical and chemical properties of edamame during bean development and application of spectroscopy-based machine learning methods to predict optimal harvest time.

This study aims to investigate the changes in physical and chemical properties of edamame during bean development and apply a spectroscopy-based machine learning (ML) technique to determine optimal harvest time. The edamame harvested at R5 (beginning seed), R6 (full seed), and R7 (beginning maturity) growth stages were characterized for physical and chemical properties, and pods were measured for spectral reflectance (360-740 nm) using a handheld spectrophotometer. The samples were categorized into 'early', 'ready', and 'late' based on the characterized properties. The results showed that pod/bean weight and pod thickness peaked at R6 and remained stable thereafter. Sugar, starch, alanine, and glycine also peaked at R6 but proceeded to decline. The ML method (random forest classification) using pods' spectral reflectance had a high accuracy of 0.95 for classifying 'early' and 'late' samples and 0.87 for classifying 'early' and 'ready' samples. Therefore, this method can determine the optimal harvest time of edamame.

Genome-wide association analysis of sucrose and alanine contents in edamame beans.

The sucrose and Alanine (Ala) content in edamame beans significantly impacts the sweetness flavor of edamame-derived products as an important attribute to consumers' acceptance. Unlike grain-type soybeans, edamame beans are harvested as fresh beans at the R6 to R7 growth stages when beans are filled 80-90% of the pod capacity. The genetic basis of sucrose and Ala contents in fresh edamame beans may differ from those in dry seeds. To date, there is no report on the genetic basis of sucrose and Ala contents in the edamame beans. In this study, a genome-wide association study was conducted to identify single nucleotide polymorphisms (SNPs) related to sucrose and Ala levels in edamame beans using an association mapping panel of 189 edamame accessions genotyped with a SoySNP50K BeadChip. A total of 43 and 25 SNPs was associated with sucrose content and Ala content in the edamame beans, respectively. Four genes (Glyma.10g270800, Glyma.08g137500, Glyma.10g268500, and Glyma.18g193600) with known effects on the process of sucrose biosynthesis and 37 novel sucrose-related genes were characterized. Three genes (Gm17g070500, Glyma.14g201100 and Glyma.18g269600) with likely relevant effects in regulating Ala content and 22 novel Ala-related genes were identified. In addition, by summarizing the phenotypic data of edamame beans from three locations in two years, three PI accessions (PI 532469, PI 243551, and PI 407748) were selected as the high sucrose and high Ala parental lines for the perspective breeding of sweet edamame varieties. Thus, the beneficial alleles, candidate genes, and selected PI accessions identified in this study will be fundamental to develop edamame varieties with improved consumers' acceptance, and eventually promote edamame production as a specialty crop in the United States.

Analysis of Shoot Architecture Traits in Edamame Reveals Potential Strategies to Improve Harvest Efficiency.

Edamame is a type of green, vegetable soybean and improving shoot architecture traits for edamame is important for breeding of high-yield varieties by decreasing potential loss due to harvesting. In this study, we use digital imaging technology and computer vision algorithms to characterize major traits of shoot architecture for edamame. Using a population of edamame PIs, we seek to identify underlying genetic control of different shoot architecture traits. We found significant variations in the shoot architecture of the edamame lines including long-skinny and candle stick-like structures. To quantify the similarity and differences of branching patterns between these edamame varieties, we applied a topological measurement called persistent homology. Persistent homology uses algebraic geometry algorithms to measure the structural similarities between complex shapes. We found intriguing relationships between the topological features of branching networks and pod numbers in our plant population, suggesting combination of multiple topological features contribute to the overall pod numbers on a plant. We also identified potential candidate genes including a lateral organ boundary gene family protein and a MADS-box gene that are associated with the pod numbers. This research provides insight into the genetic regulation of shoot architecture traits and can be used to further develop edamame varieties that are better adapted to mechanical harvesting.