Detection of Asymptomatic Candidatus Liberibacter asiaticus-Infected Citrus using an Electronic Nose

Anthoni Sulthan Harahap, Prof. Dr. Ir. Siti Subandiyah, M.Agr.Sc.

2026 | Disertasi | S3 Ilmu Pertanian

Huanglongbing (HLB) merupakan penyakit jeruk yang paling merusak di seluruh dunia. Penelitian ini menggunakan hidung elektronik portabel dengan 10 sensor metal-oxide semiconductor (MOS) untuk infeksi Candidatus Liberibacter asiaticus (CLas) yang asimptomatik pada daun jeruk Siam Purworejo. Sebanyak 454 sampel daun dikonfirmasi menggunakan polymerase chain reaction (PCR) untuk menentukan kelompok sampel sehat dan kelompok terinfeksi CLas asimptomatik. Penelitian diawali dengan menentukan teknik preparasi sampel yang paling optimal, dilanjutkan penggunaan hidung elektronik dengan optimasi windowing dan penyesuaian hiperparameter untuk menguji kemampuan hidung elektronik dalam mendeteksi infeksi CLas asimptomatik. Kinerja hidung elektronik divalidasi dengan analisis headspace gas chromatography–mass spectrometry (HS-GC/MS) terhadap profil komponen organik volatil (KOV) yang dilanjutkan dengan analisis kemometrik untuk mengetahui kandidat biomarker. Preparasi sampel optimal menggunakan daun utuh dengan proses pemanasan. Kinerja terbaik e-nose dengan optimasi windowing diperoleh pada model quadratic discriminant analysis (QDA) tiga window, dengan akurasi cross-validation (CV) 81,99 ± 4,24% dan validasi eksternal 81,00%. Kinerja terbaik hidung elektronik dengan optimasi hyperparameter tuning diperoleh pada model extra trees dengan akurasi latihan 84,57% (CI: 95% [CI]: 80,98–88,15%), akurasi uji 79.00%, spesifisitas 67.00%, dan sensitivitas 91.00%. Analisis HS-GC/MS mengonfirmasi perbedaan komposisi KOV antara kedua kelompok sampel. Meskipun principal component analysis (PCA) menunjukkan keragaman didalam grup yang substansial, analisis tersupervisi mengidentifikasi tiga putatif biomarker terpenoid yaitu 1-methyl-4-(1-methylethylidene)-cyclohexene, (+)-4-carene, dan ?-myrcene. Panel optimal yang terdiri atas 10 senyawa menghasilkan AUC sebesar 0,96 (CI 95%: 0,23–1,00) dan akurasi 88% menggunakan MCCV PLS-DA.

Huanglongbing (HLB) is one of the most destructive citrus diseases worldwide. This study used a portable, laboratory-built electronic nose (e-nose) equipped with 10 metal-oxide semiconductor (MOS) sensors for detection of asymptomatic Candidatus Liberibacter asiaticus (CLas) infection in Purworejo Siamese citrus leaves. A total of 454 leaf samples were confirmed by polymerase chain reaction (PCR) to establish the healthy sample group and the asymptomatic CLas-infected groups. The study first identified the most suitable sample preparation method, then used the e-nose to convert volatile organic compound (VOC) interactions into electrical signals for pre-processing, feature extraction, and machine-learning classification. Model optimisation were conducted using windowing and hyperparameter tuning. The e-nose performance was corroborated using headspace gas chromatography–mass spectrometry (HS-GC/MS) analysis of VOC profiles, followed by chemometric analysis. The optimal sample preparation method was whole leaves with heating. The best e-nose performance The best e-nose performance was obtained with a 3 windowing quadratic discriminant analysis (QDA) model, achieving 81.99±4.24% cross-validation accuracy (CV) and 81.00% external validation accuracy. The best e-nose performance under hyperparameter tuning was obtained with a stratified five-fold cross-validated extra trees model achieved 84.57% training accuracy (95% confidence interval [CI]: 80.98–88.15%), test accuracy 79.00%, specificity 67.00%, and sensitivity 91.00%. HS-GC/MS confirmed differences in VO,C composition between the two groups of sample. Although principal component analysis (PCA) showed substantial within-group variability, supervised analysis identified three terpenoid putative biomarkers: 1-methyl-4-(1-methylethylidene)-cyclohexene, (+)-4-carene, and ?-myrcene. The optimal 10-compound panel achieved an area under the curve (AUC) of 0.96 (95% CI: 0.23–1.00) and 88?curacy using Monte Carlo cross-validation partial least squares-discriminant analysis (MCCV PLS-DA).

 

Kata Kunci : Huanglongbing; e-nose; asymptomatic infection; volatile organic compounds; headspace gas chromatography–mass spectrometry