Scholarlink

<p>Weather based models are central to mycotoxin risk assessment, but field level discrimination can be limited when environmental training data are sparse and models must transfer to locations not used during training. We tested whether wheat ear microbiome state provides a transferable biological indicator for within season classification of trichothecene contamination under strict spatial holdout, with relevance for food and feed safety. Across three winter wheat trial locations within a single production year, ears were sampled at BBCH 77 to 83, corresponding to mid to late grain filling, and paired bacterial 16&#xa0;S rRNA gene and fungal ITS profiles were analysed alongside LC-MS/MS measurements of DON and T-2 + HT-2. Under leave one location out evaluation, microbiome informed models achieved substantially higher discrimination of higher versus lower contamination than station level weather summaries under identical spatial holdout. Weather features added little in the single season dataset, but simulation analyses indicated that environmental predictors become more learnable with temporal replication, supporting a future two stage framework in which early season weather context is refined by within season microbiome state. Portability screening and drop one ablation prioritised microbial indicators with sign consistent associations across locations and non-redundant predictive value. Together, these findings provide proof of concept that wheat ear microbiome state can support spatially transferable within season mycotoxin risk classification and targeted surveillance when environmental training data are limited.</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

Wheat Ear Microbiome State as a Transferable Indicator of Within-Season Mycotoxin Risk

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年7月24日

查看全文
  • Oluwatobi Kolawole,
  • Anne D. van Diepeningen,
  • Josipa Grzetic Martens,
  • Ane Arrizabalaga-Larrañaga,
  • Lisa Black,
  • Leo van Overbeek

摘要

Weather based models are central to mycotoxin risk assessment, but field level discrimination can be limited when environmental training data are sparse and models must transfer to locations not used during training. We tested whether wheat ear microbiome state provides a transferable biological indicator for within season classification of trichothecene contamination under strict spatial holdout, with relevance for food and feed safety. Across three winter wheat trial locations within a single production year, ears were sampled at BBCH 77 to 83, corresponding to mid to late grain filling, and paired bacterial 16 S rRNA gene and fungal ITS profiles were analysed alongside LC-MS/MS measurements of DON and T-2 + HT-2. Under leave one location out evaluation, microbiome informed models achieved substantially higher discrimination of higher versus lower contamination than station level weather summaries under identical spatial holdout. Weather features added little in the single season dataset, but simulation analyses indicated that environmental predictors become more learnable with temporal replication, supporting a future two stage framework in which early season weather context is refined by within season microbiome state. Portability screening and drop one ablation prioritised microbial indicators with sign consistent associations across locations and non-redundant predictive value. Together, these findings provide proof of concept that wheat ear microbiome state can support spatially transferable within season mycotoxin risk classification and targeted surveillance when environmental training data are limited.