Network complexity and deterministic assembly: How plant residues reshape microbial communities for pathogen suppression in degraded soils

Intensive agricultural practices can lead to the degradation of soil microbial communities while promoting the proliferation of pathogenic bacteria. Although it is known that the addition of plant residues can improve soil quality, the mechanisms through which these residues alter microbial communities and suppress diseases have not been fully elucidated. This study applies fermented residues from four medicinal plants ( Schisandra chinensis [WW], Epimedium brevicornu [YY], Rubia cordifolia [QC], and Medicago sativa [MX]) to soil where ginseng has been cultivated for an extended period. By measuring soil properties and enzyme activities, and integrating high-throughput sequencing, machine learning, and structural equation modeling, we systematically evaluated the changes in microbial community networks and key species. The results indicate that residues from WW and YY have the most significant effects—reducing the abundance of the fungal pathogen Fusarium by 63.16 % and increasing the activity of the antioxidant enzyme SOD by 88.04 %. These residues shifted the assembly mechanism of microbial communities from stochastic to deterministic (R² of the neutral model for bacteria: 0.651), enriched key beneficial species (such as Sphingomonas and Hypomyces ), and made the microbial network more complex (with a 38.6 % increase in network complexity for the WW treatment). Random forest analysis revealed that each residue treatment corresponded to unique indicator species (e.g., Dyadobacterium in the MX treatment and Leucosporidium in the WW treatment), which were closely associated with soil fertility (acid phosphatase [ACP] and organic matter [OM]) and the suppression of pathogenic bacteria. Further confirmation through Partial Least Squares Path Modeling demonstrated that the residues directly inhibited Fusarium by altering soil properties and key species. In conclusion, medicinal plant residues, especially those from WW and YY, can reconstruct soil microbial networks and enhance beneficial species, providing targeted regulatory strategies for soil restoration and the control of soil-borne diseases in intensive agriculture. • Medicinal plant residue input shifts microbial assembly from stochastic to deterministic (NCM R 2 0.079 to 0.125), enriching keystone taxa (e.g., Gemmatimonadaceae ) that suppress pathogens. • The cascading effects of residue type, soil properties, microbial biomarkers, and pathogen suppression were quantified using structural equation modeling; alfalfa residue reduced Fusarium abundance by 66.65% through increasing soil pH and enriching Actinobacteria. • Integrated random forest and network analysis identified cross-domain microbial biomarkers (e.g., Dyadobacter and Leucosporidium ) predictive of soil health restoration, which showed significant positive correlations with soil enzyme activities and provide early diagnostic targets for disease control.