Scholarlink

<p>Conventional screening of sugarcane varieties for diseases is time-consuming, tedious, and sometimes unreliable due to environmental influence on symptom expression. Marker-assisted selection can help in screening parental genotypes, but it requires the availability of one or a few large-effect molecular markers. Genomic selection, on the other hand, shows significant promise in predicting the disease response of sugarcane genotypes based on their genome-wide marker information. Three non-parametric machine learning models, support vector machine (SVM), random forest (RF), and neural network (NN) were used on historical non-continuous data to predict reactions of sugarcane genotypes to smut, leaf scald, and mosaic disease. SVM edged slightly over RF with better predictive ability (PA) for smut and mosaic resistance with ordinal data but both models performed equally for categorical data. Contrarily, RF outperformed SVM and NN with the highest PA for leaf scald with both ordinal and categorical data. This study underscores the potential of ML in disease response prediction that can be integrated into the genomic selection scheme to select and breed improved sugarcane varieties with enhanced disease resistance.</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 Enables Genomic Prediction of Resistance to Sugarcane (Saccharum spp. Hybrids) Diseases in Louisiana

  • 期刊论文

  • 发表日期：

    2025年8月13日

查看全文
  • Dipendra Shahi,
  • Jeffrey W. Hoy,
  • Anna L. Hale,
  • Brayden A. Blanchard,
  • Andre B. Gama,
  • Niranjan Baisakh

摘要

Conventional screening of sugarcane varieties for diseases is time-consuming, tedious, and sometimes unreliable due to environmental influence on symptom expression. Marker-assisted selection can help in screening parental genotypes, but it requires the availability of one or a few large-effect molecular markers. Genomic selection, on the other hand, shows significant promise in predicting the disease response of sugarcane genotypes based on their genome-wide marker information. Three non-parametric machine learning models, support vector machine (SVM), random forest (RF), and neural network (NN) were used on historical non-continuous data to predict reactions of sugarcane genotypes to smut, leaf scald, and mosaic disease. SVM edged slightly over RF with better predictive ability (PA) for smut and mosaic resistance with ordinal data but both models performed equally for categorical data. Contrarily, RF outperformed SVM and NN with the highest PA for leaf scald with both ordinal and categorical data. This study underscores the potential of ML in disease response prediction that can be integrated into the genomic selection scheme to select and breed improved sugarcane varieties with enhanced disease resistance.