Location: Subtropical Plant Pathology ResearchTitle: Evaluation of Quantitative Real-Time PCR Assays for Detection of Citrus Greening) Author
|Shatters, Robert - Bob|
Submitted to: Workshop Proceedings
Publication Type: Proceedings
Publication Acceptance Date: 6/5/2009
Publication Date: 6/5/2009
Citation: Turechek, W., Gottwald, T.R., Hartung, J.S., Hilf, M.E., Keremane, M.L., Lin, H., Shatters, R.G., Irey, M., Sieburth, P., Brlansky, R., Dagraca, J., Graham, J., Kunta, M., Roberts, P., Rogers, M., Sun, X., Wang, N. 2009. Evaluation of Quantitative Real-Time PCR Assays for Detection of Citrus Greening. Workshop Proceedings. 158-160. Interpretive Summary:
Technical Abstract: Citrus huanglongbing (HLB), or citrus greening, is a serious and industry-limiting disease. Preliminary diagnoses can be made through visual symptoms, and greater certainty can be achieved through quantitative real-time PCR (qPCR). Several qPCR procedures are available including those by designed by Li et al., Wang et al., Shatters et al., and by Irey. It is not clear if any of these tests perform better than the others under any given set of conditions. Evaluating diagnostic tests typically relies on the ability to pre-classify diseased and healthy samples with a ‘gold standard’ or reference test so that the sensitivity and specificity of the new test can be accurately estimated. However, no gold standard test exists for HLB so the objective of this study was to evaluate the performance of several qPCR protocols as used in several diagnostic laboratories using methods specific for evaluating diagnostic tests when a gold standard is absent. A total of 276 DNA samples (extracts) were sent for analysis to 13 labs in FL, CA, TX, and MD, but only results from 10 of the labs were used for analysis. The DNA samples consisted of presumed positive and negative samples, samples that tested as questionable in preliminary testing, and water blanks containing no DNA. qPCR procedures varied among labs, but all samples were run blind by all labs. For the primers developed by Li et al., the mean Ct value was used and was calculated from 8 of the labs. The primers of Wang et al., Irey, and Shatters et al. were run in only one, but different labs. In performing the analysis, the qPCR data were partitioned into two groups: test negative samples were samples that exceeded a predetermined threshold Ct value; test positive samples were samples that did not exceed the threshold Ct value. Since the threshold Ct value that best classifies samples is unknown, analyses were conducted individually for threshold values from 25 to 40. A complex multinomial model was fit to the data to estimate the sensitivity and specificity for each test at each threshold. The sensitivities and specificities were high for threshold Ct values above 32 and below 36, respectively. The Irey primers had the highest sensitivity over the greatest range of Ct values, while the Li primers had the highest specificity. The best thresholds, estimated by Youden’s index, were obtained at Ct values of 31 and 32 for all primers except for the Li primers. Youden’t index obtained its maximum at Ct = 36 for the Li primers. Results from this study indicate that greater sensitivity can be achieved with minimal loss in specificity by choosing Ct = 31 or 32 for the Wang, Irey and Shatters primers and Ct = 36 for the Li primers.