Scholarlink

<p>Fragrant pears (“<i>Pyrus sinkiangensis</i> Yu”) are deeply loved by consumers due to the delightful aroma and taste. However, fraudulent practices concerning both geographical origins and quality grades driven by economic interests have emerged. The objective of this work was to simultaneously identify geographical origins and quality grades in fragrant pears using machine learning assisted characteristic volatile organic compounds (VOCs) selection. Headspace solid-phase microextraction combined with gas chromatography-mass spectrometry (HS-SPME-GC-MS) successfully qualified and quantified the 35 VOCs in fragrant pears from different geographical origins and quality grades. Subsequently, a machine learning workflow was constructed for simultaneous geographical origins and quality grade identification. Compared to variable importance of projection (VIP) of partial least squares-discriminant analysis (PLS-DA), Boruta algorithm (BA) enhanced the characteristic VOCs selection and exhibited the more effective model performance improvement for both geographical origins and quality grades. The XGB model with BA characteristic VOCs exhibited the best generalization performance for both geographical origins and quality grades, achieving the accuracies of 100.0% and 90.0%, respectively. Among the characteristic VOCs, Hexanal and à-Farnesene exhibited the highest importance, which may become the potential biomarkers for geographical origin and quality grade identification, respectively. Overall, machine learning presents a promising tool to assist characteristic VOCs selection for simultaneously identifying the geographical origins and quality grades in fragrant pears, ensuring the fruit authenticity and avoiding fruit fraud.</p> Graphical Abstract <p></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 assisted characteristic volatile organic compounds selection for simultaneous geographical origin and quality grade identification in fragrant Pears

  • 期刊论文

  • 发表日期：

    2026年3月3日

查看全文
  • Tao Jiang,
  • Jian Yang,
  • Yan Liu,
  • Weidan Zuo,
  • Shaofeng Yuan,
  • Hang Yu,
  • Zhigang Liu,
  • Weirong Yao

摘要

Fragrant pears (“Pyrus sinkiangensis Yu”) are deeply loved by consumers due to the delightful aroma and taste. However, fraudulent practices concerning both geographical origins and quality grades driven by economic interests have emerged. The objective of this work was to simultaneously identify geographical origins and quality grades in fragrant pears using machine learning assisted characteristic volatile organic compounds (VOCs) selection. Headspace solid-phase microextraction combined with gas chromatography-mass spectrometry (HS-SPME-GC-MS) successfully qualified and quantified the 35 VOCs in fragrant pears from different geographical origins and quality grades. Subsequently, a machine learning workflow was constructed for simultaneous geographical origins and quality grade identification. Compared to variable importance of projection (VIP) of partial least squares-discriminant analysis (PLS-DA), Boruta algorithm (BA) enhanced the characteristic VOCs selection and exhibited the more effective model performance improvement for both geographical origins and quality grades. The XGB model with BA characteristic VOCs exhibited the best generalization performance for both geographical origins and quality grades, achieving the accuracies of 100.0% and 90.0%, respectively. Among the characteristic VOCs, Hexanal and à-Farnesene exhibited the highest importance, which may become the potential biomarkers for geographical origin and quality grade identification, respectively. Overall, machine learning presents a promising tool to assist characteristic VOCs selection for simultaneously identifying the geographical origins and quality grades in fragrant pears, ensuring the fruit authenticity and avoiding fruit fraud.

Graphical Abstract