Alteration of the Soil Microbiota in Ginseng Rusty Roots: Application of Machine Learning Algorithm to Explore Potential Biomarkers for Diagnostic and Predictive Analytics

Conceptualization to utilize microbial composition as a prediction tool has been widely applied in human cohorts, yet the potential capacity of soil microbiota as a diagnostic tool to predict plant phenotype remains unknown. Here, we collected 130 soil samples which are 54 healthy controls and 76 ginseng rusty roots (GRRs). Alpha diversities including Shannon, Simpson, Chao1, and phylogenetic diversity were significantly decreased in GRR (P < 0.05). Moreover, we identified 30 potential biomarkers. The optimized markers were obtained through fivefold cross-validation on a support vector machine and yielded a robust area under the curve of 0.856. Notably, evaluation of multi-index classification performance including accuracy, F1-score, and Kappa coefficient also showed robust discriminative capability (90.99%, 0.903, and 0.808). Taken together, our results suggest that the disease affects the microbial community and offers the potential ability of soil microbiota to identifying farms at the risk of GRR.