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

Abstract

Early accurate detection of crop disease is extremely important for timely disease management. Huanglongbing (HLB), one of the most destructive citrus diseases, has brought about severe economic losses for the global citrus industry. The direct strategies for HLB identification, such as quantitative real-time polymerase chain reaction (qPCR) and chemical staining, are robust for the symptomatic plants but powerless for the asymptomatic ones at the early stage of affection. Thus, it is very necessary to develop a practical method used for the early detection of HLB. In this study, a novel method combining ultra-high performance liquid chromatography/mass spectrometry (UHPLC/MS)-based nontargeted metabolomics and machine learning (ML) was developed for conducting the early detection of HLB for the first time. Six ML algorithms were selected to build the classifiers. Regularized logistic regression (LR-L2) and gradient-boosted decision tree (GBDT) outperformed with the highest average accuracy of 95.83% to not only classify healthy and infected plants but identify significant features. The proposed method proved to be practical for early detection of HLB, which tackled the shortcomings of low sensitivity in the conventional methods and avoid the problems such as lighting condition interference in spectrum/image recognition-based ML methods. Additionally, the discovered biomarkers were verified by the metabolic pathway analysis and content change analysis, which was remarkably consistent with the previous reports.

References

[1] Graca JV . Citrus greening disease. Annu Rev Phytopathol. 1991; 29: 109-36.

[2] Jagoueix S, Bove JM, Garnier M . The phloem-limited bacterium of greening disease of citrus is a member of the alpha subdivision of the Proteobacteria. Int J Syst Bacteriol. 1994; 44: 379-86.

[3] Texeira DC, Ayres J, Kitajima EW et al. First report of a Huanglongbing-like disease of citrus in Sao Paulo state, Brazil and association of a new Liberibacter species, “ Candidatus Liberibacter americanus”, with the disease . Plant Dis. 2005; 89: 107-7.

[4] Kramer J, Simnitt S & Calvin L . Fruit and Tree Nuts Outlook: March 2020. 2020. https://www.ers.usda.gov/publications/pub-details/?pubid=98169.

[5] USDA. Citrus: world markets and trade. 2020. https://www.fas.usda.gov/data/citrus-world-markets-and-trade.

[6] Bové JM. Huanglongbing: a destructive, newly-emerging, century-old disease of citrus. J Plant Pathol. 2006; 88: 7-37.

[7] Tsai JH, Liu YH . Biology of Diaphorina citri (Homoptera: Psyllidae) on four host plants. J Econ Entomol. 2000; 93: 1721-5.

[8] Roistacher CN . (eds) Graft-Transmissible Diseases of Citrus: Handbook for Detection and Diagnosis (Food and Agriculture Organization of the United Nations, 1991).

[9] Li WB, Hartung JS, Levy L . Quantitative real-time PCR for detection and identification of Candidatus Liberibacter species associated with citrus huanglongbing. J Microbiol Methods. 2006; 66: 104-15.

[10] Pandey SS, Wang N . Targeted early detection of citrus Huanglongbing causal agent ’ Candidatus Liberibacter asiaticus’ before symptom expression . Phytopathology. 2019; 109: 952-9.

[11] Lan YB, Huang ZX, Deng XL et al. Comparison of machine learning methods for citrus greening detection on UAV multispectral images. Comput Electron Agric. 2020; 171: 105234.

[12] Sankaran S, Ehsani R, Etxeberria E . Mid-infrared spectroscopy for detection of Huanglongbing (greening) in citrus leaves. Talanta. 2010; 83: 574-81.

[13] Sankaran S, Mishra A, Maja JM et al. Visible-near infrared spectroscopy for detection of Huanglongbing in citrus orchards. Comput Electron Agric. 2011; 77: 127-34.

[14] Wetterich CB, De Oliveira F, Neves R et al. Detection of Huanglongbing in Florida using fluorescence imaging spectroscopy and machine-learning methods. Appl Opt. 2016; 56: 15-23.

[15] Sanchez L, Pant S, Mandadi K et al. Raman spectroscopy vs quantitative polymerase chain reaction in early stage Huanglongbing diagnostics. Sci Rep. 2020; 10: 10101.

[16] Yao MY, Fu GR, Xu J et al. In situ diagnosis of mature HLB-asymptomatic citrus fruits by laser-induced breakdown spectroscopy. Appl Opt. 2021; 60: 5846-53.

[17] Albrecht U, Fiehn O, Bowman KD . Metabolic variations in different citrus rootstock cultivars associated with different responses to Huanglongbing. Physiologie végétale. 2016; 107: 33-44.

[18] Fiehn O . Metabolomics-the link between genotypes and phenotypes. Plant Mol Biol. 2002; 48: 155-71.

[19] Suh JH, Guha A, Wang ZX et al. Metabolomic analysis elucidates how shade conditions ameliorate the deleterious effects of greening (Huanglongbing) disease in citrus. Plant J. 2021; 108: 1798-814.

[20] Cajka T, Fiehn O . Toward merging untargeted and targeted methods in mass spectrometry-based metabolomics and lipidomics. Anal Chem. 2016; 88: 524-45.

[21] Raftery D, ed. Mass Spectrometry in Metabolomics. Springer: New York; 2014.

[22] Perez De Souza L, Alseekh S, Naake T et al. Mass spectrometry-based untargeted plant metabolomics. Curr Protoc Plant Biol. 2019; 4: e20100.

[23] Bzdok D, Altman N, Krzywinski M . Statistics versus machine learning. Nat Methods. 2018; 15: 233-4.

[24] Ananthakrishnan G, Choudhary N, Roy A et al. Development of primers and probes for genus and species specific detection of ‘ Candidatus Liberibacter species’ by real-time PCR . Plant Dis. 2013; 97: 1235-43.

[25] Rstudio Team . RStudio: integrated development environment for R. 2021. http://www.rstudio.com.

[26] R Core Team . R: A Language and Environment for Statistical Computing. 2020. http://www.r-project.org/index.html.

[27] Qin J, Lou YF . L1-2 Regularized Logistic Regression, in Proceedings of the 53rd Asilomar Conference on Signals, Systems, and Computers 779-783. United States: Pacific Grove; 2019.

[28] Basu S, Kumbier K, Brown JB et al. Iterative random forests to discover predictive and stable high-order interactions. P Natl A Sci. 2018; 115: 1943-8.

[29] Friedman JH . Stochastic gradient boosting. Comput Stat Data Anal. 2002; 38: 367-78.

[30] Shan J, Wang YY, Gao W . Prediction of chemical exergy of organic substances using artificial neural network-multi layer perceptron. Energy Sources, Part A: Recovery, Utilization, and Environmental Effects. 2018; 40: 1826-32.

[31] Krstajic D, Buturovic LJ, Leahy DE et al. Cross-validation pitfalls when selecting and assessing regression and classification models. J Cheminformatics. 2014; 6: 10.

[32] Van Rossum G, Drake FL Jr, eds. Python Tutoria Centrum voor Wiskunde en Informatica. Centrum voor Wiskunde en Informatica Amsterdam: Amsterdam; 1995.

[33] Sokolova M, Japkowicz N, Szpakowicz S . Beyond Accuracy, F-Score and ROC: A Family of Discriminant Measures for Performance Evaluation. In: AI 2006: Advances in Artificial Intelligence.Springer Berlin Heidelberg: Berlin, Heidelberg, 2006.

[34] Boughorbel S, Jarray F, El-Anbari M . Optimal classifier for imbalanced data using Matthews correlation coefficient metric. PLoS One. 2017; 12: e0177678.

[35] Marrocco C, Duin RPW, Tortorella F . Maximizing the area under the ROC curve by pairwise feature combination. Pattern Recogn. 2008; 41: 1961-74.

[36] Wang Z, Li J, Chambers A et al. Rapid structure-based annotation and profiling of dihydrochalcones in star fruit (Averrhoa carambola) using UHPLC/Q-Orbitrap-MS and molecular networking . J Agric Food Chem. 2021; 69: 555-67.

[37] Picart-Armada S, Fernández-Albert F, Vinaixa M et al. FELLA: an R package to enrich metabolomics data. BMC Bioinformatics. 2018; 19: 538.

[38] Kanehisa M, Goto S . KEGG: Kyoto encyclopedia of genes and genomes. Nucleic Acids Res. 2000; 28: 27-30.

[39] Peng T, Kang JL, Xiong XT et al. Integrated transcriptomics and metabolomics analyses provide insights into the response of Chongyi wild mandarin to Candidatus Liberibacter asiaticus infection . Front Plant Sci. 2021; 12: 748209.

[40] Wei X, Mira A, Yu QB et al. The mechanism of citrus host defense response repression at early stages of infection by feeding of Diaphorina citri transmitting Candidatus Liberibacter asiaticus. Front Plant Sci. 2021; 12: 635153.

[41] Killiny N, Nehela Y . Metabolomic response to Huanglongbing: role of carboxylic compounds in Citrus sinensis response to ’ Candidatus Liberibacter asiaticus’ and its vector . MPMI. 2017; 30: 666-78.

[42] Hung WL, Wang Y . A targeted mass spectrometry-based metabolomics approach toward the understanding of host responses to Huanglongbing disease. J Agric Food Chem. 2018; 66: 10651-61.

[43] Ma WX, Pang ZQ, Huang XE et al. Citrus Huanglongbing is a pathogen-triggered immune disease that can be mitigated with antioxidants and gibberellin. Nat Commun. 2022; 13: 529.

[44] Hijaz FM, Manthey JA, Folimonova SY et al. An HPLC-MS characterization of the changes in sweet orange leaf metabolite profile following infection by the bacterial pathogen Candidatus Liberibacter asiaticus . PLoS One. 2013; 8: e79485.

[45] Xue AH, Liu YQ, Li HX et al. Early detection of Huanglongbing with EESI-MS indicates a role of phenylpropanoid pathway in citrus. Anal Biochem. 2021; 639: 114511.

[46] Indrakumari R, Poongodi T, Khaitan S et al. A review on plant diseases recognition through deep learning, in Handbook of Deep Learning in Biomedical Engineering 1st edn. In: Balas VE, Mishra BK, Kumar R, eds. Elsevier Science: Amsterdam, 2021.

[47] Ferentinos KP . Deep learning models for plant disease detection and diagnosis. Comput Electron Agric. 2018; 145: 311-8.

[48] Ma JC, Du KM, Zheng FX et al. A recognition method for cucumber diseases using leaf symptom images based on deep convolutional neural network. Comput Electron Agric. 2018; 154: 18-24.

[49] Schumann A, Waldo L, Mungofa P et al. Computer tools for diagnosing citrus leaf symptoms (part 2): smartphone apps for expert diagnosis of citrus leaf symptoms. EDIS. 2020; 2020: SL478.

[50] Mujika KM, JaJ M, De Miguel AF . Advantages and disadvantages in image processing with free software in radiology. J Med Syst. 2018; 42: 36.