Title: Detecting cotton boll rot with an electronic nose Authors
Submitted to: Journal of Cotton Science
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: June 2, 2014
Publication Date: N/A
Interpretive Summary: Standard methods of detecting diseased plants and identifying the pathogen(s) responsible for the disease require traditional isolation and culturing techniques which can be tedious and time-consuming. We examined the use of a commercially-available electronic nose (e-nose) to detect cotton fruit (bolls) infected with the bacterium Pantoea agglomerans, the main causative agent of the emerging cotton disease known as South Carolina Boll Rot. We also examined the ability of the e-nose to discriminate between the odors emitted by liquid cultures of P. agglomerans and Klebsiella pneumoniae, another opportunistic bacterium of cotton bolls. Our results indicate the two species of bacteria emit different odors and bolls infected with P. agglomerans emit an odor which is different than that of healthy bolls. However, the e-nose could not accurately discriminate between the odors produced by each species of bacteria or between the odors produced by infected and healthy bolls. In light of the dismal performance of the e-nose, our results suggest this particular e-nose does not provide the level of discrimination accuracy needed in crop disease management.
Technical Abstract: South Carolina Boll Rot is an emerging disease of cotton, Gossypium hirsutum L., caused by the opportunistic bacteria, Pantoea agglomerans (Ewing and Fife). Unlike typical fungal diseases, bolls infected with P. agglomerans continue to appear normal externally, complicating early and rapid detection of diseased bolls. We examined the use of a commercially-available electronic nose (e-nose) to distinguish between liquid cultures of P. agglomerans and another opportunistic bacterial pathogen of bolls, Klebsiella pneumoniae (Schroeter). We also examined whether the e-nose could accurately discriminate between P. agglomerans-infected and non-infected bolls at one and two weeks post-infection. The e-nose was trained to recognize headspace collections of volatiles emitted from treatments established in each experiment. Cross-validation of the training data sets indicated the smellprints of the Luria Bertani liquid medium and each species of bacteria cultured in the medium could be discriminated with 69% accuracy. However, upon testing samples of each treatment solution, only 49% of the samples were correctly identified. In the second experiment, cross-validation of the training set indicated the smellprints of P. agglomerans-infected and non-infected bolls at one and two weeks post-infection could be distinguished with 63% accuracy. However, upon testing the discrimination accuracy of the e-nose, <30% of the test bolls were correctly classified. In light of the dismal to marginal performance of the Cyranose 320 in our experiments, the discrimination accuracy of this particular e-nose needs to be vastly improved before it can be recommended and adopted as a crop disease management tool.