Modeling the Accuracy of Xylella fastidiosa Molecular Diagnostic Tests in Naturally Infected Almond Tree Samples

Xylella fastidiosa (Xf) is a quarantine plant pathogen in the European Union, recognized as a high-priority pest due to its devastating cultural and economic impact on crops, ornamental plants, and landscape vegetation. The development and implementation of reliable, sensitive, and specific diagnostic methods for Xf detection are critical to ensure the production and trade of healthy plant material and to facilitate effective control measures, primarily aimed at eradication. Despite the availability of numerous detection protocols, their diagnostic parameters remain not precisely defined, and no universally accepted gold standard protocol exists. This study compared the global accuracy and performance of six molecular assays using almond samples collected from naturally infected almond trees in the Alicante Demarcated Area, Spain. Additionally, the study evaluated the influence of plant sample type (leaf petioles versus woody chips) on diagnostic accuracy. Harper-qPCR and Li-qPCR assays demonstrated the highest sensitivity, with detection limits as low as 2.8 to 3 fg of Xf DNA. Droplet digital PCR (ddPCR) exhibited excellent sensitivity for woody chip samples, while Li-qPCR showed superior specificity across both tissue types. In contrast, recombinase polymerase amplification displayed lower detection limits and reproducibility compared with qPCR-based methods. Bayesian latent class models indicated that combining Harper-qPCR and Li-qPCR for petioles, or Harper-qPCR and ddPCR for wood samples, optimized diagnostic reliability by reducing false negatives, which is critical in buffer zones under eradication while maintaining high specificity. These findings emphasize the need for tailoring diagnostic protocols to the epidemiological context, balancing sensitivity and specificity to optimize surveillance schemes for Xf and to support effective phytosanitary management strategies.[Formula: see text] Copyright © 2026 The Author(s). This is an open access article distributed under the CC BY-NC-ND 4.0 International license.