Agricultural pesticides are significant drivers of antibiotic resistance in soil. However, the differential impacts of antibiotic versus non-antibiotic pesticides on the soil resistome are poorly characterized. Here, we analyzed sequencing data from soils exposed to either antibiotic or non-antibiotic pesticides to compare differences in antibiotic resistance gene (ARG) burden, diversity, assembly processes, network topology, and host taxonomy. Soils exposed to antibiotic pesticides exhibited a significantly higher ARG burden (0.52% vs. 0.27% of total genes), whereas soils exposed to non-antibiotic pesticides showed significantly higher alpha diversity (p < 0.05). ARG community compositions also differed significantly between antibiotic and non-antibiotic exposures (PERMANOVA, R² = 0.215, p < 0.001). Assembly analysis using the modified stochasticity ratio indicated that deterministic processes governed ARG community assembly in both groups, with stronger influence observed in non-antibiotic soils. Co-occurrence network analysis revealed contrasting patterns. A compact, highly centralized network emerged in antibiotic-exposed soils, while a larger, more dispersed network characterized non-antibiotic soils. In both networks, aminoglycoside ARGs served as keystone nodes, accompanied by the β-lactam ARG in antibiotic soils and the macrolide ARG in non-antibiotic soils. Pseudomonadota was the predominant ARG host (>60% contribution) across both exposures, though many other phyla exhibited significance (p < 0.05) between group differences in their ARG contributions. Non-pathogenic bacteria comprised the majority of ARG hosts in all samples. When examining ARG contributions from pathogenic hosts, zoonotic and animal-associated pathogens contributed significantly (p < 0.01) more in non-antibiotic soils than in antibiotic soils, whereas the ARG contribution from plant pathogens was comparable between the two pesticide groups. Overall, our study suggests that antibiotic and non-antibiotic pesticides shape distinct ARG network patterns and host–pathogen profiles, posing distinct risks to public health and agricultural ecosystems.

Untargeted metabolomics based on ultra-high performance liquid chromatography-tandem mass spectrometry (UPLC-MS/MS) was used to explore metabolite differences in the peel and flesh tissues of three Rong’an kumquat cultivars. A total of 5043 metabolites were detected. Principal component analysis (PCA) revealed clear separation of metabolic profiles among different cultivars. Significantly altered metabolites were screened with thresholds of VIP > 1 and |Log₂FC| ≥ 1.0. A total of 193 flavonoid metabolites were annotated, among which 84 flavones and 28 flavonols were the dominant subclasses. Flavonoid profiles exhibited significant cultivar- and tissue-specific differences. Huapi kumquat (HP) and Cuimi kumquat (CM) showed tissue-specific anthocyanin accumulation, with anthocyanins enriched mainly in peel tissues, while Youpi kumquat (YP) accumulated abundant chalcones and flavanones. The flavonoid metabolic variation in peel tissues was greater than that in flesh tissues. 2′,6′-dihydroxy-4,4′-dimethoxychalcone (with log₂FC up to 8.44) and xanthohumol were identified as candidate metabolic markers. KEGG pathway enrichment analysis further revealed distinct tissue-biased metabolic features. Specifically, cultivar-related metabolic variation in peel was associated with diterpenoid and lipid metabolic pathways, whereas flesh metabolic divergence was linked to indole alkaloid biosynthesis, histidine metabolism, and folate biosynthesis. This study systematically elucidated the metabolic characteristics and flavonoid differences in the peel and flesh of three kumquat cultivars, providing candidate metabolic markers and theoretical support for the evaluation and utilization of kumquat germplasm resources.