Scholarlink

<p>Understanding the genetic control of biomass yield and phenolic composition in sorghum (<i>Sorghum bicolor</i> L.) is crucial for optimizing bioenergy and biomaterial applications. We analyzed publicly available phenotypic and genotypic (192,040 SNPs) data from a diverse panel of 96 sorghum genotypes, including radiation-induced mutants. We compared single-SNP linear regression (RS) with machine learning (ML) models (Bootstrap Forest, BF; Boosted Tree, BT) to identify SNPs associated with four agronomic traits and seven phenolic compound measurements. RS analysis yielded significant SNPs (FDR &lt; 0.01) only for luteolinidin diglucoside. Conversely, the ML models prioritized numerous SNPs across most traits based on importance scores. To understand the collective biological function of these loci, a Gene Ontology (GO) enrichment analysis was performed. This functional analysis revealed a novel and consistent enrichment for processes related to transposable element (TE) activity, such as DNA replication and modification, across both agronomic and phenolic traits. This suggests that TE-induced variation, likely activated by mutagenesis, is a major source of the observed phenotypes in this population. We hypothesize that ML models excel at identifying the complex genetic signals of TE-mediated effects that are often missed by conventional linear models. Integrating ML-based SNP importance ranking with functional genomics offers a promising strategy for refining and identifying novel candidate loci. These findings not only provide a valuable set of candidates for functional validation and targeted breeding efforts but also highlight the critical role of TEs in generating agronomically important variation in mutagenized germplasm.</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

Machine learning reveals signatures of transposable element activity driving agronomic and phenolic traits in mutagenized sorghum

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年9月10日

查看全文
  • Ezekiel Ahn,
  • Sookyung Oh,
  • Jacob Botkin,
  • Jishnu Bhatt,
  • Dongho Lee,
  • Clint Magill

摘要

Understanding the genetic control of biomass yield and phenolic composition in sorghum (Sorghum bicolor L.) is crucial for optimizing bioenergy and biomaterial applications. We analyzed publicly available phenotypic and genotypic (192,040 SNPs) data from a diverse panel of 96 sorghum genotypes, including radiation-induced mutants. We compared single-SNP linear regression (RS) with machine learning (ML) models (Bootstrap Forest, BF; Boosted Tree, BT) to identify SNPs associated with four agronomic traits and seven phenolic compound measurements. RS analysis yielded significant SNPs (FDR < 0.01) only for luteolinidin diglucoside. Conversely, the ML models prioritized numerous SNPs across most traits based on importance scores. To understand the collective biological function of these loci, a Gene Ontology (GO) enrichment analysis was performed. This functional analysis revealed a novel and consistent enrichment for processes related to transposable element (TE) activity, such as DNA replication and modification, across both agronomic and phenolic traits. This suggests that TE-induced variation, likely activated by mutagenesis, is a major source of the observed phenotypes in this population. We hypothesize that ML models excel at identifying the complex genetic signals of TE-mediated effects that are often missed by conventional linear models. Integrating ML-based SNP importance ranking with functional genomics offers a promising strategy for refining and identifying novel candidate loci. These findings not only provide a valuable set of candidates for functional validation and targeted breeding efforts but also highlight the critical role of TEs in generating agronomically important variation in mutagenized germplasm.