Portable HS-SPME-GC-MS combined with machine learning for onsite detection of huanglongbing-infected navel orange leaves at high-altitude environments

Abstract

Huanglongbing (HLB) remains one of the most destructive diseases threatening global citrus production, with its impact exacerbated in diverse altitudinal environments where environmental conditions can influence disease progression and diagnostic accuracy. In this study, we applied headspace solid-phase microextraction (HS-SPME) coupled with portable gas chromatography-mass spectrometry (GC–MS) to analyze volatile profiles of orange leaves from healthy and asymptomatic HLB-infected trees at different altitudes. A comprehensive machine learning (ML) framework, comprising random forest, logistic regression, XGBoost, support vector machine, and Ensemble classifier, was then used to achieve accurate discrimination between infection states and to identify altitude-dependent shifts in key metabolites such as limonene, 3-carene, and citronellal. The results revealed that the abundance of HLB-infected metabolites varied with altitude, indicating that environmental factors should be considered when selecting robust biomarkers for disease diagnosis. This portable analytical platform enables rapid and reliable detection of HLB under varying environmental conditions, providing a practical tool for precision agriculture and advancing the understanding of citrus metabolic responses to biotic (HLB) and abiotic (altitude) stresses.