Multispectral imaging and automated analysis for quantifying grain quality to reveal known and potential novel alleles affecting grain traits in wheatThe assessment of wheat seed quality and underlying genetics using multispectral imaging and automated trait analysis based on seed-level morphological and spectral properties

Ji Zhou^1,2*Jie Dai³Daiki Abe^1,4Zhengjie Wen³Yuyi Li^1,4Hongyan Li³Jinlong Huang³Phil Howell¹Robert Jackson^1*

• ¹National Institute of Agricultural Botany (NIAB), Cambridge, United Kingdom
• ²CAS Center for Excellence in Molecular Plant Sciences, Shanghai, China
• ³Nanjing Agricultural University, Nanjing, China
• ⁴University of Cambridge, Cambridge, United Kingdom

To accelerate the pace of wheat (Triticum aestivum L.) production worldwide, seed quality and desired seed-level characteristics have received growing attention as they directly impact early seedling establishment, seed longevity, and grain quality, which are agronomically important for carrying out wheatcrop improvement programs. Nevertheless, the throughput and accuracy of seed phenotyping and automated analysis of wheat seeds lots have become a key limiting factor, which urgently requiringe new solutions. In this study, we integrated combined automated multispectral seed imaging (MSI) using; i.e. the VideometerLab 4 and Autofeeder systems) with a variety of machine learning (ML) and computer vision (CV) techniques to conduct high-throughput morphological and spectral seed-level trait analysis at the seed level. Using 493 lines selected from the NIAB Diverse MAGIC (NDM) wheat population, we first developed ML/CV combined algorithms to quantify eight morphological traits (e.g. Original Research seed size, length, width, and roundness) and eight spectral traits, ranging from ultraviolet (i.e. 375 nm, which signifying correlates with crude protein) to near-infrared (e.g. 975 nm, signifying for assessing water content) wavelengths, were quantified. Then, we applied genome-wide association studies (GWAS) to establish a phenotype-to-genotype analytic pipeline, linking automatically computationally derived measured 16 seed quality related traits to genetic loci, resulting in nine significant loci previously reported and several unknown loci that are valuable for further assessment. Taken together, we believe that this integrated multispectral MSIimaging and analyticsis platform provides a powerful route solution for seed research and seed-focused crop improvement in wheat, enabling us to bridge MSI, data-drivenseed-level trait analysis, and genetic mapping to assess seed morphology, seed quality, seed-level physiology, and their underlying genetic architectures effectively.