Scholarlink

<p>Soybean oil composition directly influences its shelf life, nutritional value, and suitability for diverse cooking and biodiesel production applications. Several traits underlying oil composition, particularly fatty acid profiles, are complex and costly to phenotype. In this study, we assessed the genomic prediction accuracy for total oil and five principal fatty acids in soybean using two genotyping platforms, DNA arrays (<i>BarcSoySNP6k</i>) and Genotyping-by-Sequencing (GBS), in combination with different quality control (QC) parameters. We also compared two Bayesian models: BayesB and Bayesian Ridge Regression (BRR). While both platforms generally yielded moderate to high prediction accuracies, the DNA array consistently provided stable results across traits and QC thresholds. In contrast, GBS performance was more variable but superior for stearic acid, likely due to its broader allele discovery. The choice of prediction model had minimal effect on accuracy, whereas filtering thresholds, especially missing data cutoffs, significantly influenced results in GBS data. Among all filter combinations, the 5% minor allele frequency, 20% heterozygotes, and 30% missing data thresholds delivered the highest average accuracies and are recommended for similar studies. Strong correlations among fatty acids suggest opportunities for indirect selection strategies, particularly for traits like linolenic acid, which correlate with total oil content, an accessible trait to phenotype. This study provides a precise and reproducible strategy for improving genomic prediction accuracy in soybean breeding through informed choices of genotyping and filtering for genomic selection.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Impact of genotyping platform and quality control on genomic prediction accuracy of soybean oil composition

  • 期刊论文

  • 发表日期：

    2025年12月9日

查看全文
  • Melina Prado,
  • Regina Helena Geribello Priolli,
  • Evellyn Giselly Oliveira Couto,
  • Felipe Sabadin,
  • Kaio Olimpio das Graças Dias,
  • José Baldin Pinheiro

摘要

Soybean oil composition directly influences its shelf life, nutritional value, and suitability for diverse cooking and biodiesel production applications. Several traits underlying oil composition, particularly fatty acid profiles, are complex and costly to phenotype. In this study, we assessed the genomic prediction accuracy for total oil and five principal fatty acids in soybean using two genotyping platforms, DNA arrays (BarcSoySNP6k) and Genotyping-by-Sequencing (GBS), in combination with different quality control (QC) parameters. We also compared two Bayesian models: BayesB and Bayesian Ridge Regression (BRR). While both platforms generally yielded moderate to high prediction accuracies, the DNA array consistently provided stable results across traits and QC thresholds. In contrast, GBS performance was more variable but superior for stearic acid, likely due to its broader allele discovery. The choice of prediction model had minimal effect on accuracy, whereas filtering thresholds, especially missing data cutoffs, significantly influenced results in GBS data. Among all filter combinations, the 5% minor allele frequency, 20% heterozygotes, and 30% missing data thresholds delivered the highest average accuracies and are recommended for similar studies. Strong correlations among fatty acids suggest opportunities for indirect selection strategies, particularly for traits like linolenic acid, which correlate with total oil content, an accessible trait to phenotype. This study provides a precise and reproducible strategy for improving genomic prediction accuracy in soybean breeding through informed choices of genotyping and filtering for genomic selection.