Abstract Soil microbiota is central to agroecosystem sustainability, influencing fertility, plant health, and resilience. Agricultural practices shape these communities in distinct ways: conventional systems, with intensive chemical inputs, often promote microbial homogeneity, whereas organic systems, with higher organic matter inputs and reduced disturbance, tend to sustain functionally diverse microbiomes. Yet, the extent of these effects remains poorly understood, particularly when taxonomic, functional, and machine-learning–based indicators are evaluated jointly across different cropping systems. To address this gap, we compared the taxonomic and functional diversity of soil microbiota in organic and conventional fields cultivated with common bean ( Phaseolus vulgaris ) and grapevine ( Vitis vinifera ) in Brazil. Microbial communities were profiled using 16 S rRNA and ITS sequencing, with analyses spanning diversity metrics, taxonomic composition, functional inference, and machine-learning-based biomarker identification. Bacterial diversity was greater under conventional management, while fungal diversity did not differ between systems. Machine-learning approaches identified Bradyrhizobium and Roseococcus (bacteria) and Lecythophora (fungus) as consistent biomarkers of organic soils. Functional predictions pointed to enhanced nitrogen fixation and carbon cycling in organic fields, whereas sulfur cycling was more prominent in conventional ones. Microbial community structure clearly separated by management and correlated strongly with soil chemistry, and Venn analysis revealed more unique bacterial and fungal operational taxonomic units (OTUs) in organic soils. Together, these results highlight how agricultural management acts as a strong ecological filter of soil microbiota. Conventional farming favored broader bacterial diversity. In contrast, organic systems fostered distinct microbial biomarkers and enhanced functional potential, particularly for nitrogen and carbon pathways.
Kawakami et al. (Thu,) studied this question.