Distinctive bioactive compound and endophytic microbiota profiles of Glycyrrhiza and Astragalus: Machine learning-assisted biomarker discovery and classification

Glycyrrhiza and Astragalus are representative industrial crops within the Fabaceae family and important bioresource plants widely cultivated across northern China. However, the regional variation patterns underlying their bioactive compound composition and endophytic microbiome characteristics remain insufficiently understood. In this study, we systematically analyzed 120 samples collected from eight major production regions using targeted quantification of bioactive compounds and 16S rRNA and ITS high-throughput sequencing to investigate endophytic microbial community dynamics in two representative Fabaceae crops. Our results demonstrated clear host-specific divergence in endophytic microbiota between Glycyrrhiza and Astragalus . Samples from the Zhangxian and Longxi production regions in Gansu exhibited elevated levels of Liquiritin and Glycyrrhizic acid and showed distinct microbial community composition patterns, including enrichment of Bacteroides and depletion of Pseudomonas . Compared with Glycyrrhiza , Astragalus harbored significantly higher endophytic alpha diversity and exhibited greater fungal community dispersion. Bacterial communities showed stronger responsiveness to regional variation in bioactive compound levels than fungal communities, indicating differential ecological sensitivity between microbial kingdoms. Co-occurrence network analysis further revealed distinct regional microbial association patterns. The Gansu regions of Zhangxian and Longxi displayed microbial networks with higher modularity, whereas the Baotou production region in Inner Mongolia exhibited a network structure with higher connectivity. These contrasting network topologies indicate region-associated differences in microbial community organization within Fabaceae crop systems. Based on microbial taxonomic and functional signatures, we further established 11 machine learning models that accurately classified crop genera, geographic origins, and groups categorized by bioactive compound content, achieving high discriminatory performance (AUC > 0.90). Collectively, this study provides insights into regional variation of endophytic microbiomes in industrial Fabaceae crops and highlights the potential of endophytic microbiome-derived signatures for crop classification, geographic traceability, ecological characterization, and bioactive compound-associated quality evaluation.