Seed coat color genetics and genotype x environment effects in yellow beans via machine-learning and genome-wide association

Authors: SADOHARA, RIE - Michigan State University; LONG, YUNFEI - Michigan State University; IZQUIERDO, PAULO - Michigan State University; URREA, CARLOS - University Of Nebraska; MORRIS, DANIEL - Michigan State University; Cichy, Karen

Submitted to: The Plant Genome
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 9/22/2021
Publication Date: 11/24/2021
Citation: Sadohara, R., Long, Y., Izquierdo, P., Urrea, C., Morris, D., Cichy, K.A. 2021. Seed coat color genetics and genotype x environment effects in yellow beans via machine-learning and genome-wide association. The Plant Genome. 15(1). Article e20173. https://doi.org/10.1002/tpg2.20173.
DOI: https://doi.org/10.1002/tpg2.20173

Interpretive Summary: The seed color of dry beans is diverse with various colors, patterns, and tendencies to darken postharvest. Bean breeders develop varieties with color characteristics that meet the market demands in their target regions. Characterization of seed coat color is typically conducted with color charts or a colorimeter. The challenge with the colorimeter is that it includes the entire area of the seed, which may not be uniformly colored and therefore give inaccurate color readings. The goal of this research was to develop an image and machine learning based seed color evaluation method that removes background hilum and reflection to provide a more accurate representation of the seed color. This method was applied to measure the seed color of a set of 295 bean lines grown in two locations (Michigan and Nebraska) for two field seasons. The information was used to understand the genetic and environmental control of color and seed coat postharvest darkening. Genomic regions associated with color and darkening were identified. Environmental variability was also detected such that seeds of beans grown in Michigan tended to be darker than those grown in Nebraska. The machine learning based color quantification tool and the related genomic information for seed coat color are useful breeding and research tools directly applicable to meet consumers’ expectations for bean seed appearance.

Technical Abstract: Common bean (Phaseolus vulgaris L.) is consumed worldwide, and regional preferences exist for seed characteristics, including color, size, and shape. Colors of the seed coat, hilum ring, and corona are all important, along with susceptibility to postharvest darkening, which decreases seed value. This study aimed to characterize a collection of 295 yellow bean genotypes for seed color and postharvest darkening behavior, evaluate genotype × environment effects on them and map those traits via genome-wide association analysis. The yellow beans were grown for two years in Michigan and Nebraska, USA, and were evaluated for L*a*b*, postharvest darkening, and hilum ring and corona colors. A model to exclude the hilum ring and corona of the seeds, black background, and light reflection was developed by using machine learning, allowing for targeted and efficient L*a*b* value extraction from the seed coat. The genotype × environment effects were significant for the color values, and MI-grown seeds had darker seeds than NE-grown seeds. SNPs were associated with L* and hilum ring color on Pv10 near the J gene. An SNP on Pv07 associated with L*, a*, postharvest darkening, and hilum ring and corona colors was near the P, the ground factor gene for seed coat color expression. The machine learning-aided model to extract color values from the seed coat, the wide variability in seed morphology traits and the associated SNPs are useful breeding and research tools directly applicable to meet consumers’ expectations for bean seed appearance.