Early detection of Huanglongbing using mass spectrometry-based proteomics and machine learning J. MOHR (1), J. Chavez (2), J. Ramsey (3), J. Mahoney (4), T. Thannhauser (5), K. Howe (5), M. Alexander (1), K. Godfrey (6), E. Chin (6), C. Slupsky (6), J. Bruce (2), M. Cilia (5). (1) Cornell University, Ithaca, NY, U.S.A.; (2) University of Washington, Seattle, WA, U.S.A.; (3) Boyce Thompson Institute, Ithaca, NY, U.S.A.; (4) Boyce Thompson Institute, Ihtaca, NY, U.S.A.; (5) USDA-ARS, Ithaca, NY, U.S.A.; (6) UC Davis, Davis, CA, U
Huanglongbing (HLB) is the most serious disease of citrus and is associated with infection by <i>Candidatus </i>Liberibacter asiaticus (CLas). Typically, infected trees have no visible symptoms for the first few months, making it difficult to correctly diagnose infected trees. Current detection strategies rely on real-time, quantitative PCR for detection of CLas 16s rDNA, but the success of this approach depends on the presence of a sufficiently high titer of CLas in the host tissue sampled. Measuring systemic changes in plant proteins in response to HLB offers an opportunity for early detection of HLB infection that does not rely on direct detection of the pathogen. Using tandem mass tag labeling and high resolution mass spectrometry, we constructed proteomic profiles of HLB-positive Lisbon Lemon at two-week intervals over the first several months of infection. Samples from the same set of trees were analyzed by RNAseq to generate an annotated database of expressed genes and predicted proteins to facilitate analysis of proteomic data. Proteins differentially expressed between control and infected samples were identified at each time point, and evaluated for their use as potential biomarkers. A drastic change in protein response between control and infected lemon trees was measured at 10-weeks post-grafting. A support vector machine model was generated and trained using selected differentially expressed proteins, which may improve early detection in field samples. View Presentation |