Virome datasets and viral metagenome-assembled genomes from aquaculture-impacted freshwater environments Authors: Kira Moon, Ilnam Kang, Jang-Cheon Cho Published: 2026 ABSTRACT Bacteriophages in natural environments play a critical role in microbial ecology by regulating bacterial populations, mediating nutrient cycling, and facilitating horizontal gene transfer. Aquaculture operations, particularly inland fish farms, are major sources of anthropogenic influence on freshwater ecosystems. Here, we present three viral metagenomic datasets derived from freshwater samples collected at an inland aquaculture effluent site and adjacent upstream and downstream locations along the Sung-am River in Jincheon County, South Korea. The datasets were generated using the Illumina HiSeq X sequencing platform, yielding approximately 10.0–11.2 Gbp per sample. Quality assessments confirmed minimal bacterial contamination, with negligible proportions of rRNA and bacterial marker genes. Assembly using metaSPAdes and MEGAHIT, application of Phables to resolve high-quality phage genomes (viral metagenome-assembled genomes; vMAGs), viral identification with VirSorter2, and clustering using Vclust, resulted in 2,837–3,156 virus operational taxonomic units (vOTUs; ≥10 kb) per sample. Each vOTU sequence is analyzed for taxonomic assignment and putative host prediction. These datasets provide a valuable resource for further studies on viral diversity and microbial ecology in freshwater ecosystems affected by aquaculture. Background & Summary Bacteriophages (phages), despite being among the smallest biological entities in natural ecosystems, outnumber all other life forms and play critical ecological roles1. Particularly, phages significantly influence microbial population dynamics by infecting and lysing bacterial cells, thereby regulating bacterial abundance and community structure. Through lytic activity, phages release intracellular contents into the surrounding environment, consequently altering local nutrient concentrations and contributing to nutrient cycling processes2,3. In addition, phages are well-established mediators of horizontal gene transfer (HGT), facilitating the exchange of genetic material among bacterial populations, which enhances bacterial genetic diversity4. Bacterial metabolic genes that are transferred among bacterial populations by phages are referred to as auxiliary metabolic genes (AMGs). Frequently found within phage genomes, AMGs can significantly influence the metabolism of their bacterial hosts5,6. AMGs encompass a broad array of metabolic functions, including pathways involved in carbohydrate, lipid, and amino acid metabolism, as well as genes associated with pathogenicity and antibiotic resistance7. These genes can transiently enhance host metabolism during phage infection, ultimately benefiting the propagation of phage progeny. HGT mediated by phages contributes substantially to the genetic diversity and adaptive capacity of microbial communities, particularly through lysogenic infections5,8. Recent advancements in viral metagenomics (virome analysis) have facilitated the discovery of diverse AMGs across various environmental viral communities9–11. Virome analysis, which enables the direct recovery of viral contigs and viral metagenome-assembled genomes (vMAGs) without the need for cultivation, offers a valuable strategy for identifying and tracking phage-associated AMGs. Among the genes transferred by phages, antibiotic resistance genes (ARGs) are particularly noteworthy in the context of ARG dissemination12–14. Although phages themselves are not affected by antibiotics, they can inadvertently acquire ARGs from bacterial hosts and subsequently transfer these genes to other bacteria, potentially increasing antibiotic resistance within microbial communities. Identifying and cataloging ARGs from virome datasets not only expands ARG databases, traditionally derived from bacterial genomes, but also supports the development of new antibiotics and resistance mitigation strategies. Given the importance of virome analysis for characterizing phage-mediated AMG transfer, in this study, we investigated metagenomes of phage populations and the distribution of phage-encoded ARGs in aquaculture-impacted freshwater environments. Aquaculture operations often involve the intensive use of nutrient-rich feed and antibiotics to promote fish growth and prevent disease. A significant proportion of these inputs remain in the rearing water as uneaten feed, fish excreta, and residual antibiotic compounds15. Because aquaculture effluent is frequently discharged directly into nearby rivers with minimal wastewater treatment, it can serve as a direct human impact on natural freshwater ecosystems, introducing elevated nutrient loads and antibiotics that can shape microbial and viral communities16. Surface water samples were collected from an inland carp aquaculture effluent (AQ-EF), as well as from upstream (AQ-UP) and downstream (AQ-DW) sites along the Sung-am River in Jincheon County, South Korea (Fig. 1). By generating viral metagenomic datasets from these samples and analyzing the assembled viral contigs and vMAGs for ARG content, we aimed to contribute to the expansion of public ARG databases and provide a valuable resource for future antimicrobial resistance surveillance.Fig. 1Sampling sites and workflow of viral metagenome preparation and data processing. (a) Map of Sung-am river in Jincheon County, South Korea, showing the three sampling sites: AQ-UP (upstream), AQ-EF (aquaculture effluent discharge point), and AQ-DW (downstream). (b) Schematic workflow illustrating the main steps involved in viral metagenome preparation, sequencing, and bioinformatic analysis. Sampling sites and workflow of viral metagenome preparation and data processing. (a) Map of Sung-am river in Jincheon County, South Korea, showing the three sampling sites: AQ-UP (upstream), AQ-EF (aquaculture effluent discharge point), and AQ-DW (downstream). (b) Schematic workflow illustrating the main steps involved in viral metagenome preparation, sequencing, and bioinformatic analysis. The three virome samples were sequenced using the Illumina HiSeq X platform, generating approximately 10.0–11.2 Gbp of data per sample, corresponding to 66–74 million reads (Table 1). Quality assessment using ViromeQC17 indicated minimal contamination, as evidenced by extremely low proportions of reads aligned to SSU and LSU rRNA genes (~0.001%) and bacterial marker genes (~0.01%), together with enrichment scores of 58.5–87.7 (Table 1) that support the purity and integrity of the viral datasets. However, taxonomic classification using Kraken218, with the option of “–confidence 0.1” to reduce spurious assignments, revealed that a large proportion of reads (78.4–81.6%) remained unclassified, with only 4.3–5.1% assigned as viral in origin (Fig. 2). This observation is common in virome studies and primarily reflects the limited representation of viral sequences in currently available reference databases. Sequencing reads were assembled into contigs using metaSPAdes v4.0.019 and MEGAHIT v1.2.920. The resulting assembly graphs were used as input for Phables21 to construct high-quality phage genomes (viral metagenome-assembled genomes; vMAGs). Contigs generated by metaSPAdes and MEGAHIT, and vMAGs produced by Phables were pooled, and a length threshold of ≥ 10 kb was applied. Sequences meeting this criterion were curated using VirSorter222 to identify viral sequences, followed by clustering into vOTUs with Vclust23 based on the MIUViG24 thresholds. This workflow resulted in 3,156 vOTUs for AQ-UP, 2,890 for AQ-EF, and 2,837 for AQ-DW. The quality and taxonomy information of all representative sequences of these vOTUs (hereafter, vOTUs) are provided in Table S1.Table 1Overview of sampling sites, sequencing yield, GC content, and viral enrichment validation for the three freshwater virome datasets.AQ-UPAQ-EFAQ-DWGPS location36.7833 N, 127.4330 E36.7665 N, 127.4439 E36.7603 N, 127.4435 ESampling date2019-04- 232019-04-232019-04-23Sequencing platformIllumina HiSeqIllumina HiSeqIllumina HiSeqBase pair (Gbp)11.1710.2910.00GC content %43.247.344.6Q20 (%)95.4894.7897.29Q30 (%)89.5587.8992.69ViromeQC % of SSU rRNA0.00150.00100.0013 % of LSU rRNA0.00110.00100.0014 % of bacterial markers0.01200.00800.0083 Score58.587.784.6NCBI accession No.*SAMN46862211SAMN46862212SAMN46862213SRR32380190JBMFGI000000000SRR32380189JBMFGH000000000SRR32380188JBMFGG000000000*NCBI accession numbers are provided for three record types: SAMN (BioSample), SRR (Sequence Read Archive), and JBMFGI/JBMFGH/JBMFGG (GenBank WGS, BioProject PRJNA1224589).Fig. 2Taxonomic composition of viral metagenomes based on Kraken2 analysis using the core_nt database. The pie charts display the proportion of reads classified at the domain level, while the Sankey diagrams illustrate the taxonomic assignments of viral reads across different taxonomic ranks. Results are shown for (a) AQ-UP, (b) AQ-EF, and (c) AQ-DW. Abbreviations for the labels on the horizontal axes: D, Domain; P, Phylum; C, Class; O, Order; F, Family; G, Genus; S, Species. Table 1Overview of sampling sites, sequencing yield, GC content, and viral enrichment validation for the three freshwater virome datasets.AQ-UPAQ-EFAQ-DWGPS location36.7833 N, 127.4330 E36.7665 N, 127.4439 E36.7603 N, 127.4435 ESampling date2019-04- 232019-04-232019-04-23Sequencing platformIllumina HiSeqIllumina HiSeqIllumina HiSeqBase pair (Gbp)11.1710.2910.00GC content %43.247.344.6Q20 (%)95.4894.7897.29Q30 (%)89.5587.8992.69ViromeQC % of SSU rRNA0.00150.00100.0013 % of LSU rRNA0.00110.00100.0014 % of bacterial markers0.01200.00800.0083 Score58.587.784.6NCBI accession No.*SAMN46862211SAMN46862212SAMN46862213SRR32380190JBMFGI000000000SRR32380189JBMFGH000000000SRR32380188JBMFGG000000000*NCBI accession numbers are provided for three record types: SAMN (BioSample), SRR (Sequence Read Archive), and JBMFGI/JBMFGH/JBMFGG (GenBank WGS, BioProject PRJNA1224589). Overview of sampling sites, sequencing yield, GC content, and viral enrichment validation for the three freshwater virome datasets. SRR32380190 JBMFGI000000000 SRR32380189 JBMFGH000000000 SRR32380188 JBMFGG000000000 *NCBI accession numbers are provided for three record types: SAMN (BioSample), SRR (Sequence Read Archive), and JBMFGI/JBMFGH/JBMFGG (GenBank WGS, BioProject PRJNA1224589). Taxonomic composition of viral metagenomes based on Kraken2 analysis using the core_nt database. The pie charts display the proportion of reads classified at the domain level, while the Sankey diagrams illustrate the taxonomic assignments of viral reads across different taxonomic ranks. Results are shown for (a) AQ-UP, (b) AQ-EF, and (c) AQ-DW. Abbreviations for the labels on the horizontal axes: D, Domain; P, Phylum; C, Class; O, Order; F, Family; G, Genus; S, Species. Taxonomic identification of the vOTUs was limited due to the underrepresentation of closely related sequences in current public databases. Most of vOTUs were classified as Caudoviricetes without assignment to order or lower level of taxonomic ranks by geNomad25 (Table S1). Nevertheless, vConTACT226 clustering analyses revealed distinct, environment-specific viral communities, with many vOTUs forming co-clustered groups. Some vOTUs were affiliated with known Caudoviricetes and Flavobacterium-infecting phages, allowing for indirect phylogenetic inference (Fig. 3). Putative host predictions using iPHoP27 showed that while some vOTUs were predicted to infect archaeal hosts, the majority were associated with bacterial hosts, particularly members of the family Flavobacteriaceae (Fig. 4).Fig. 3vOTU taxonomic network based on vConTACT2 clustering. Nodes representing reference genomes from the NCBI RefSeq database are shown as circles, while vOTUs from each dataset are represented using distinct shapes: V-shaped nodes for AQ-UP, triangles for AQ-EF, and squares for AQ-DW. Node colors represent vConTACT2-derived cluster assignments. Only network clusters containing more than 20 nodes are shown.Fig. 4Predicted bacterial hosts of vOTUs inferred using the iPHoP program. Each horizontal line represents a vOTU, with the first column indicating the virome sample ID, and the second, third, and fourth columns showing the predicted bacterial host phylum, class, and order, respectively. Sankey diagrams illustrate the taxonomic flow of host predictions from phylum to order for (a) AQ-UP, (b) AQ-EF, and (c) AQ-DW. The number of vOTUs assigned to each taxonomic group is shown in parentheses. vOTU taxonomic network based on vConTACT2 clustering. Nodes representing reference genomes from the NCBI RefSeq database are shown as circles, while vOTUs from each dataset are represented using distinct shapes: V-shaped nodes for AQ-UP, triangles for AQ-EF, and squares for AQ-DW. Node colors represent vConTACT2-derived cluster assignments. Only network clusters containing more than 20 nodes are shown. Predicted bacterial hosts of vOTUs inferred using the iPHoP program. Each horizontal line represents a vOTU, with the first column indicating the virome sample ID, and the second, third, and fourth columns showing the predicted bacterial host phylum, class, and order, respectively. Sankey diagrams illustrate the taxonomic flow of host predictions from phylum to order for (a) AQ-UP, (b) AQ-EF, and (c) AQ-DW. The number of vOTUs assigned to each taxonomic group is shown in parentheses. Screening of the vOTUs for ARGs using BLASTx search against the Comprehensive Antibiotic Resistance Database28 (CARD; e-value ≤ 10−4, bitscore ≥ 100, percent identity ≥ 40%) identified a total of 37 ARGs across the three virome datasets (Table 3). Although diverse ARGs targeting multiple antibiotic classes, such as diaminopyrimidine antibiotics, chloramphenicol, fluoroquinolones, para-aminosalicylic acid, phenicol antibiotics, tetracycline, and virginiamycin A-like antibiotics were detected, their overall abundance within the viral fraction was extremely low. Specifically, 13 ARGs were identified in AQ-UP, 17 in AQ-EF, and 7 in AQ-DW, with Mtub_thyA_PAS being the most frequently detected gene. This suggests that while aquaculture effluent may introduce a variety of resistance genes into freshwater environments, their prevalence within the viral fraction remains limited. While the overall abundance of ARGs was low, the detailed cataloging of these genes presented herein contributes valuable information to public ARG databases and may provide a foundation for future research on ARG dynamics. Importantly, beyond ARG detection, this study provides virome datasets from aquaculture-impacted freshwater environments, which remain underrepresented in viral metagenomic research. As such, the datasets presented here contribute to expanding current knowledge of aquatic viral communities and offer an important resource for studying virus-host interactions, viral diversity, and the ecological impacts of aquaculture. Methods Environmental sampling and viral metagenome sequencing In April 2019, surface freshwater samples were collected from three sites along the Sung-am River in Jincheon County, South Korea (Fig. 1). The sampling sites included the effluent discharge point (AQ-EF) of an inland carp aquaculture farm, located just before the effluent enters to the main river channel, a site approximately 2 km upstream (AQ-UP), and a site 700 m downstream (AQ-DW) from the discharge point. Approximately 4 L of surface water samples were collected at each location in sterile water bottles, maintained at 4°C, and promptly transported to the laboratory. To concentrate viral particles, the collected water samples were first filtered through 0.2 μm pore-size Supor® PES membrane filters (Pall Corporation, New York, USA) to remove bacterial cells. Ferric chloride (0.1 mg/L) was added to the filtrate, followed by vigorous shaking and incubation in the dark at room temperature for at least 1 hour to induce flocculation of viral particles via FeCl3 association29. Approximately 4 L of treated water samples were then filtered through 148-mm diameter, 0.8 μm Isopore® polycarbonate filters (Merck Millipore, Darmstadt, Germany) to collect the flocculated viral particles. The virus-containing filters were stored in the dark at 4°C until further processing. Physicochemical parameters of the water samples were measured on-site using a YSI 556 MPS Multiprobe System (YSI Inc., Yellow Springs, OH, USA). For viral nucleic acid extraction, viral particles were resuspended from the polycarbonate filters using an ascorbic acid buffer (0.1 M EDTA, 0.2 M MgCl2, 0.2 M ascorbic acid; pH 6)30. The resuspended viral particles were treated with DNase I (10 U/ml) and RNase A (1 U/ml) (Sigma-Aldrich, St. Louis, MO, USA) to degrade external nucleic acids and minimize contamination in the viral metagenome. Enzyme activity was halted by adding EDTA and EGTA to a final concentration of 100 mM31. The samples were concentrated using Amicon® Ultra centrifugal filters (50 kDa MWCO; Merck, Darmstadt, Germany) to reduce the volume to approximately 100 μl. The concentrates were further sterilized by filtration through a 0.2-μm Supor® syringe filter. Viral DNA was extracted from the final filtrates using the DNeasy Blood & Tissue Kit (Qiagen, Hilde, Germany) following the manufacturer’s protocol with minor modifications. Specifically, 70 μl of the filtrate was combined with 300 μl of ATL buffer, 30 μl of Proteinase K, and 6 μl of RNase A to lyse capsid proteins and remove RNA. Subsequently, 300 μl each of ethanol and AL buffer were added. The DNA was purified using spin columns and eluted in 50 μl of EB buffer. DNA concentration and quality were assessed using the Qubit™ dsDNA High Sensitivity Assay Kit (Thermo Fisher Scientific, Waltham, MA, USA). High-quality viral DNA was used to construct TruSeq libraries (Illumina). Sequencing was performed by Macrogen (Seoul, South Korea) on the Illumina HiSeq X platform, generating 2 × 151 bp paired-end reads. An overview of the viral metagenome preparation and sequencing statistics is presented in Fig. 1 and Table 1. Quality trimming and assembly of viral metagenome reads and contigs Raw metagenomic reads were preprocessed and quality-trimmed using BBDuk from the BBTools suite with the following parameters: a k-mer length threshold of 23 bp for contaminant removal and a minimum read length of 100 bp. The resulting high-quality reads were assembled using two different assemblers to maximize assembly quality: metaSPAdes (v4.0.0)19 and MEGAHIT (v1.2.9.)20. k-mer sizes of “21, 33, 55, 77, 99, 127” and “31, 51, 71, 91, 111, 131” were used for metaSPAdes and MEGAHIT, respectively. Assembly graphs generated by metaSPAdes (“assembly_graph_after_simplification.gfa”) were used as input for Phables (v1.4.1)21 to reconstruct high-quality phage genomes (vMAGs). Assembly graphs files of METAHIT were built from respective contigs using megahit_toolkit (with the “contig2fastg 131” option), followed by conversion into gfa files using gfatools (https://github.com/lh3/gfatools; with the “view graph.fastg > graph.gfa” option), then were also used as input for Phables. Contigs obtained from the assemblers (metaSPAdes and MEGAHIT) and vMAGs resolved by Phables (“resolved_paths.fasta” from both assemblers) were subsequently pooled per each sample and filtered by length (≥10 kbp). These sequences were processed by VirSorter2 (v2.2.4)22 to identify viral sequences. All sequences identified as viral were then clustered into vOTUs using Vclust23, with the options “–algorithm leiden–metric ani–ani 0.95–qcov 0.85–out-repr” to follow MIUViG24 thresholds (ANI ≥ 95% and aligned fraction ≥ 85%) and obtain representative sequences of the vOTUs for downstream analyses. This produced 3,156 (AQ-UP), 2,890 (AQ-EF), and 2,837 (AQ-DW) vOTUs (Table 2). Additionally, CheckV (v0.7.0)32 was used to assess the completeness and potential contamination of vOTUs, but without further quality-based filtering based on the results. Quality-controlled virome reads from AQ-UP, AQ-EF, and AQ-DW were mapped to the three vOTU sets using CoverM33 to estimate the proportion of reads that are represented by the vOTUs (Table 4).Table 2Summary of sequencing output and assembly statistics for vOTUs.AQ-UPAQ-EFAQ-DWRaw reads73,987,77268,112,18466,214,818QC passed reads73,847,43268,053,54666,095,598AssemblermetaSPAdes 4.0, MEGAHIT 1.2.9metaSPAdes 4.0, MEGAHIT 1.2.9metaSPAdes 4.0, MEGAHIT 1.2.9Total assembled sequences*1,820,802 (1,290)606,745 (232)983,397 (1,543)Total assembled sequences (≥10 kb)**5,6124,8175,079Total viral assembled sequences***5,2944,5894,812vOTUs****3,1562,8902,837vMAGs among vOTUs17686255Taxonomic analysis Reads (DB)Kraken2 (core_nt, 12/28/2024)Kraken2 (core_nt, 12/28/2024)Kraken2 (core_nt, 12/28/2024) Contigs (DB)vConTACT2, (RefSeq) geNomadvConTACT2, (RefSeq) geNomadvConTACT2, (RefSeq) geNomad*The sum of the number of contigs obtained by the two assembly programs and the number of vMAGs produced by Phables. Number of vMAGs obtained by the Phables is noted in parenthesis.**The number of sequences that are equal to or longer than 10 kbp among “Total assembled sequences”***The number of sequences that are identified as viral by VirSorter2 among “Total assembled sequences (≥10 kb)”****The number of vOTUs generated by the application of Vclust to “Total viral assembled sequences”, with the MIUViG thresholds (ANI ≥ 95% and aligned fraction ≥85%).Table 3ARGs identified in vOTUs from freshwater environments using the CARD database.Target antibioticARGCARD AROAQ-UPAQ-EFAQ-DWDiaminopyrimidine antibioticsdfrA330031051ChloramphenicolPaer_catB630026781FluoroquinoloneCdif_gyrB_FLO30045622Para-aminosalicylic acidMtub_thyA_PAS300415310114Phenicol antibioticscatB3300267611catB930026811TetracyclineTxR30050082Virginiamycin A-likevatB300284112Total ARGs13177Table 4Proportion of metagenome reads mapped to vOTUs from each site.vOTU originAQ-UP virome readsAQ-EF virome readsAQ-DW virome readsAQ-UP45.9%0.8%4.7%AQ-EF2.9%73.0%15.4%AQ-DW25.7%27.3%90.8% Table 2Summary of sequencing output and assembly statistics for vOTUs.AQ-UPAQ-EFAQ-DWRaw reads73,987,77268,112,18466,214,818QC passed reads73,847,43268,053,54666,095,598AssemblermetaSPAdes 4.0, MEGAHIT 1.2.9metaSPAdes 4.0, MEGAHIT 1.2.9metaSPAdes 4.0, MEGAHIT 1.2.9Total assembled sequences*1,820,802 (1,290)606,745 (232)983,397 (1,543)Total assembled sequences (≥10 kb)**5,6124,8175,079Total viral assembled sequences***5,2944,5894,812vOTUs****3,1562,8902,837vMAGs among vOTUs17686255Taxonomic analysis Reads (DB)Kraken2 (core_nt, 12/28/2024)Kraken2 (core_nt, 12/28/2024)Kraken2 (core_nt, 12/28/2024) Contigs (DB)vConTACT2, (RefSeq) geNomadvConTACT2, (RefSeq) geNomadvConTACT2, (RefSeq) geNomad*The sum of the number of contigs obtained by the two assembly programs and the number of vMAGs produced by Phables. Number of vMAGs obtained by the Phables is noted in parenthesis.**The number of sequences that are equal to or longer than 10 kbp among “Total assembled sequences”***The number of sequences that are identified as viral by VirSorter2 among “Total assembled sequences (≥10 kb)”****The number of vOTUs generated by the application of Vclust to “Total viral assembled sequences”, with the MIUViG thresholds (ANI ≥ 95% and aligned fraction ≥85%). Summary of sequencing output and assembly statistics for vOTUs. *The sum of the number of contigs obtained by the two assembly programs and the number of vMAGs produced by Phables. Number of vMAGs obtained by the Phables is noted in parenthesis. **The number of sequences that are equal to or longer than 10 kbp among “Total assembled sequences” ***The number of sequences that are identified as viral by VirSorter2 among “Total assembled sequences (≥10 kb)” ****The number of vOTUs generated by the application of Vclust to “Total viral assembled sequences”, with the MIUViG thresholds (ANI ≥ 95% and aligned fraction ≥85%). Table 3ARGs identified in vOTUs from freshwater environments using the CARD database.Target antibioticARGCARD AROAQ-UPAQ-EFAQ-DWDiaminopyrimidine antibioticsdfrA330031051ChloramphenicolPaer_catB630026781FluoroquinoloneCdif_gyrB_FLO30045622Para-aminosalicylic acidMtub_thyA_PAS300415310114Phenicol antibioticscatB3300267611catB930026811TetracyclineTxR30050082Virginiamycin A-likevatB300284112Total ARGs13177 ARGs identified in vOTUs from freshwater environments using the CARD database. Table 4Proportion of metagenome reads mapped to vOTUs from each site.vOTU originAQ-UP virome readsAQ-EF virome readsAQ-DW virome readsAQ-UP45.9%0.8%4.7%AQ-EF2.9%73.0%15.4%AQ-DW25.7%27.3%90.8% Proportion of metagenome reads mapped to vOTUs from each site. Taxonomic assignment, host prediction, and ARG detection Taxonomic profiling of virome reads was performed using Kraken218 against the k2_core_nt_20241228 database. The resulting classifications were further refined using Bracken to obtain more accurate species- and genus-level abundances, then visualized as Sankey diagrams through the Pavian web application34 (Fig. 2). For taxonomic prediction of vOTUs, vConTACT235 was employed using the 1Jan2025_vConTACT2 database for sequence clustering. Clusters with fewer than 20 nodes or lacking nodes from our samples were excluded from visualization. The remaining clusters were visualized using Cytoscape36 (Fig. 3). We also used geNomad25 to corroborate the taxonomic assignment of vOTUs. Putative host prediction for each vOTU was performed using iPHoP27 in predict mode with the Aug_2023_pub database. The results were visualized as Sankey diagrams generated using SankeyMATIC (https://sankeymatic.com; Fig. 4). Finally, ARG screening within vOTUs was performed using BLASTx searches against the CARD database v3.3.0. Hits were filtered using an e-value (≤10−4), bitscore (≥100), and percent identity (≥40%) thresholds, and ARGs meeting these criteria are summarized in Table 3. Data Records The raw viral metagenomic sequencing data have been deposited in the NCBI Sequence Read Archive under the accession number SRP56457437. The representative sequences of the vOTUs obtained from the three samples are also available under BioProject PRJNA1224589, with accession numbers JBMFGI00000000038 (AQ-UP), JBMFGH00000000039 (AQ-EF), and JBMFGG00000000040 (AQ-DW). Technical Validation To ensure the quality and reliability of the virome datasets, sequencing quality, contamination levels, and completeness of viral genome recovery were assessed. The total base pairs obtained per sample ranged from 10.0–11.2 Gbp, with GC content between 43.2– 47.3% (Table 1). Sequencing quality was evaluated based on Q-scores, which confirmed that the majority of reads were of high quality (Table 1). Potential bacterial contamination within the virome datasets was assessed using ViromeQC17, which estimates contamination by detecting SSU and LSU rRNA genes, along with 31 conserved bacterial marker proteins. The results indicated minimal contamination, with extremely low levels of SSU and LSU rRNA genes (<0.0015%) and bacterial marker genes (0.0080–0.0120%), as well as enrichment scores of 58.5–87.7, supporting the effectiveness of the viral particle enrichment procedure employed (Table 1). Supplementary information Supplementary Table Supplementary Table