Pecans (Carya illinoinensis (Wangenh.) K. Koch) are globally consumed nuts and an important agricultural commodity in the United States. Scab is a devastating pecan disease, which necessitates the application of numerous fungicide sprays in the growing season of pecans. Even with the control measures, in wet years, scab infection results in great yield loss (over 50% loss in susceptible varieties) and deterioration of nut quality. Although there have been various efforts to alleviate the scab, the development of scab-resistant pecan cultivars is the most effective method to control the disease. However, current methods to assess pecan scab resistance require multiple years of field screening and complicated laboratory (microscopic) techniques. Thus, a simple and reliable method that can rapidly evaluate pecan scab resistance at an early stage of infection is necessary. In this study, metabolomic analysis with machine learning algorithms was utilized to identify early biomarkers for the scab resistance of pecan seedlings. Two pecan seedlings with contrasting scab resistance ('Pawnee' and 'Desirable') were inoculated with water (control), Pa-OK-11 (isolated from 'Pawnee'), and De-Tif-11 (isolated from 'Desirable') for 7 days. 'Desirable' seedlings exhibited resistance to Pa-OK-11, while 'Pawnee' seedlings showed moderate resistance to De-Tif-11. Both cultivars were susceptible to their respective isolates. Leave samples from each seedling were collected at different time points (0, 1, 2, 3, 4, 5, 7 days). For the metabolomics work, liquid chromatography‒mass spectrometry (LC‒MS) was employed to analyze metabolites in samples, which can cover a wide range of primary and secondary metabolisms, including carbon fixation, glycolysis, citric acid cycle, amino acid metabolism, phenylpropanoid, monolignol, and flavonoid biosynthesis. Different machine learning algorithms were compared to find differentially regulated metabolites (biomarkers) between scab-resistant and -susceptible seedling groups. With a combination of machine learning models, we obtained reliable potential biomarkers, e.g., phenolic acids, flavonoids, plant hormones, and their intermediates and precursors, involved in the early stage of scab infection. The selected markers are expected to be used to classify scab resistance levels in pecan seedlings within a week after infection, which may replace the conventional method (phenotype-based mass selection) for pecan breeding selection. In short, this research breaks the bottleneck of resistance screening in pecans and will help facilitate the early selection of scab-resistant pecan cultivars to achieve breeding goals.
