Correction to: Nature Communications https://doi.org/10.1038/s41467-025-57500-7, published online 05 March 2025

In the version of the article initially published, reported comparisons of the VITAP tool to PhaGCN2 were instead based on the PhaGCN tool. The article is now updated with results for PhaGCN2, affecting the text, Figs. 3a, 4a, c, Tables 1, 2, Supplementary Fig. 2a and Supplementary Data 2, 5.

In the second paragraph of the Results “Benchmarking VITAP against simulated viromes and new viruses” section, the text now reading “PhaGCN2 is unable to perform taxonomic assignments on sequences as short as 1-kb, but demonstrates comparable classification performance to other pipelines for longer sequences 5-kb” replaces the original “PhaGCN2 is unable to perform taxonomic assignments on short sequences and can only provide valid assignments for Duploviricota, its classification metrics are notably lower than those of the other pipelines.” In the fourth paragraph of the Results “Benchmarking VITAP against simulated viromes and new viruses” section, the text now reading “Meanwhile, VITAP attains annotation rates above 0.8 across 13 viral phyla on sequences of all lengths (except for Nucleocytoviricota, Artverviricota and Lenarviricota at 1 kb/5 kb). By contrast, CAT and geNomad are able to maintain annotation rates above 0.8 for all sequence lengths in only five (Peploviricota, Cressdnaviricota, Preplasmiviricota, Duplornaviricota, Kitrinoviricota) and two (Peploviricota, Uroviricota) viral phyla, respectively. PhaGCN2 achieves annotation rates of at least 0.8 for sequences longer than 5 kb from four viral phyla (Taleaviricota, Uroviricota, Saleviricota, Preplasmiviricota), whereas vConTACT2 fails to reach annotation rates of 0.8 for most viral phyla.” replaces the original “Meanwhile, VITAP attains annotation rates above 0.8 across 14 viral phyla on sequences of all lengths (except for Nucleocytoviricota and Artverviricota at 1/5 kb). By contrast, CAT and geNomad are able to maintain annotation rates above 0.8 for all sequence lengths in only three (Peploviricota, Preplasmiviricota, Duplornaviricota) and one (Uroviricota) viral phyla, respectively, whereas vConTACT2 and PhaGCN2 fail to reach annotation rates of 0.8 for most viral phyla.”

In the second paragraph of the Results “The re-assessment of viral diversity in deep-sea environments” section, the text now reading “Although VPF-Class and PhaGCN2 achieved higher genus and family-level unit assignments compared to VITAP, their results were not considered to be highly reliable and exhaustive due to their outdated databases, suboptimal taxonomic assignment efficiency or low lineage annotation (Supplementary Figs. 1 and 2)” replaces the original “Although VPF-Class and PhaGCN2 achieved higher genus and family-level unit assignments compared to VITAP, their results were not considered to be highly reliable due to their outdated databases and suboptimal taxonomic assignment efficiency (Fig. 4 and Supplementary Fig. 1).”

In the Table 1 “Short sequence (as low as 1-kb) analysis” row, as low as 1-kb replaces the original “≤5-kb,” while in Table 2, the entry for PhaGCN2; 5-kb (h), the entry now lists “9.1” h in place of the original empty cell. At the end of the Fig. 6 legend, the text now reading “…(lineage score ≥0.6) for all viral phyla, and offer acceptable classification for most viral phyla at a medium-confidence level (0.1 ≤ lineage score <0.9)” replaces the original “…(lineage score ≥0.9) for all viral phyla, and offer acceptable classification for most viral phyla at a medium-confidence level (0.3 ≤ lineage score <0.6).”

For comparison, the original Figs. 3 and 4 are available as Supplementary Information alongside this amendment. In Supplementary Fig. 2a, the PhaGCN2 results change from “6011” to “3078” for both columns. The text, figures, tables and Supplementary Fig. 2 and Supplementary Data 2, 5 are updated in the HTML and PDF versions of the article.