Non-destructive detection strategy of maize seed vigor based on seed phenotyping and the potential for accelerating breeding.

Abstract

Introduction Seeds are fundamental to agricultural production, and their vigor affects seedling quality, quantity, and crop yield. Accurate vigor assessment methods are crucial for agricultural productivity. Objectives Traditional seed vigor testing and phenotypic trait acquisition methods are complex, time-consuming, or destructive. Thus, this study aims to develop a non-destructive method for assessing maize seed vigor based on seed phenotyping and to delve into the underlying mechanism of this method. Methods Utilizing 368 maize inbred lines with diverse genetic backgrounds as research material, the cold-soaking germination percentage, closely related to the field emergence percentage, was selected to evaluate seed vigor. High and low-vigor groups were ultimately obtained through mixed grouping based on the consistent performance of seeds harvested across years. Subsequently, non-destructive techniques such as hyperspectral imaging, machine vision, and gas chromatography with ion mobility spectrometry, along with machine learning, were employed to establish models for distinguishing high and low-vigor maize seeds in their natural state. After determining the optimal strategy, key phenotypic features were identified for relevant genetic and metabolic analyses to elucidate the effectiveness of the seed vigor testing model. Results Among the evaluated methods, the machine vision-based emerged as the optimal seed vigor detection method (accuracy ≈ 90%). Subsequently, four key features (B_mean, b_mean, S_mean, and b_std) were selected for genome-wide association analysis, revealing two confident candidate genes involved in hormone regulation affecting seed germination. Further investigations confirmed significant differences in several endogenous hormones' levels and flavonoid, chlorophyll, and anthocyanidin content between high and low-vigor maize seeds. Conclusion This study validates a reliable, non-destructive seed vigor detection model supported by genetic and physiological-biochemical evidence. The findings enhance the application of non-destructive seed quality testing models and provide reliable and high-throughput measurable phenotypic traits associated with seed vigor, thereby facilitating gene mining and accelerating high-vigor maize variety breeding.

Plant phenotyping relevance

トウモロコシ種子の活力を非破壊的に推定するため、ハイパースペクトル画像、マシンビジョン、機械学習を用いた検出モデルを開発・比較・検証しており、フェノタイピング手法が中心である。

Code and data availability

The article describes seed phenotyping (hyperspectral imaging, machine vision via AIseed, GC-IMS) and machine learning/GWAS analysis, but no public phenotype dataset, seed images, analysis code, or trained model is deposited or made available. The only URL mentioned (MaizeGDB) is a generic reference genome resource, a

No evidence-backed public reproduction asset is currently recorded.