High-Throughput Seed Phenotyping and GWAS Uncover Key Genetic Variants Influencing Seed Quality in Leymus chinensis

Leymus chinensis (Trin.) Tzvel. (sheepgrass) is an important forage species, yet the relationships between seed phenotypic traits, agronomic performance, and their underlying genetic mechanisms remain unclear. In this study, we utilized the AIseed high-throughput phenotyping platform to systematically analyze 54 image-based traits (i-traits)—encompassing morphology, color, and texture—in 262 dehusked seeds of sheepgrass. Coupled with 50K single nucleotide polymorphism (SNP) chip genotyping data, we performed a genome-wide association study (GWAS) to elucidate genetic correlations among seed phenotypic traits. Elastic net regression was employed to identify informative phenotypic predictors, revealing significant associations between seed size, seed coat texture, and color with hundred-seed weight (HGW), hundred-seed weight without glumes (HGWwg), and germination rate (GR). Additionally, a germplasm screening approach based on principal component analysis (PCA) achieved a 71% accuracy rate in predicting high-germination germplasm and identified 10 germplasm lines with superior comprehensive performance. GWAS identified several SNPs significantly associated with seed color and morphology, mainly on chromosomes Lc2Xm and Lc6Xm. KEGG analysis highlighted the roles of phenylpropanoid and flavonoid biosynthesis pathways, with candidate genes such as PAL, PER18, PER50, BGLU16, BACOVA_02659, and ANR implicated. This study offers effective phenotypic screening strategies and valuable genetic resources for the molecular breeding of sheepgrass.