Nontargeted metabolomics-based multiple machine learning modeling boosts early accurate detection for citrus...

Area of Science:

• Agricultural Science
• Biotechnology
• Data Science

Background:

• Huanglongbing (HLB) is a devastating citrus disease causing significant global economic losses.
• Current detection methods like qPCR are ineffective for asymptomatic plants in early stages.
• There is a critical need for practical, early-stage HLB detection techniques.

Purpose of the Study:

• To develop and validate a novel method for early detection of Huanglongbing (HLB) in citrus plants.
• To combine ultra-high performance liquid chromatography/mass spectrometry (UHPLC/MS)-based nontargeted metabolomics with machine learning (ML) for HLB identification.
• To address the limitations of conventional methods, including low sensitivity and interference from environmental factors.

Main Methods:

• Utilized UHPLC/MS-based nontargeted metabolomics to analyze citrus plant samples.
• Applied six machine learning (ML) algorithms to build classification models for HLB detection.
• Employed regularized logistic regression (LR-L2) and gradient-boosted decision tree (GBDT) for optimal classification performance.

Main Results:

• Achieved a high average accuracy of 95.83% in classifying healthy and HLB-infected citrus plants.
• Successfully identified key metabolic features indicative of early-stage HLB infection.
• The developed ML models demonstrated superior performance compared to conventional detection strategies.

Conclusions:

• The combined metabolomics and ML approach offers a practical and sensitive method for early HLB detection.
• This novel technique overcomes the limitations of traditional methods and avoids issues like lighting interference.
• Identified biomarkers were validated through metabolic pathway and content change analyses, aligning with existing research.