Question about host information for viral sequences in the database

[ZihengWu]

Dear authors,

Thank you for developing and maintaining this very useful tool/database.

I would like to ask whether the viral sequences included in the database are associated with any host information (for example, predicted or known host taxonomy), and if so, whether this information is available to users in any form (e.g., annotation files, metadata tables, or optional outputs).

Having access to host information for viral sequences would greatly improve downstream analyses, especially for interpreting virus–host interactions and ecological patterns. If such data already exist but are not currently exposed, we would also appreciate any guidance on how to obtain or infer it.

Thank you very much for your time and for your valuable work.

Best regards

ZihengWu

[mtisza1]

Hi @ZihengWu,

This is a good question. The reason that the host information is not distributed is that I don't personally believe it would allow for accurate inference. There are three main reasons for this:

1. EsViritu takes records from GenBank, where contributors can add arbitrary (incorrect) entries in the "Host" field or leave it blank. Also, many viruses have multiple viable hosts.
2. A GenBank record reflects a single isolate (or MAG). There could be a situation where someone passages a mouse virus in a human cell line, then lists the host as Homo sapiens. This might be technically correct for that isolate but isn't broadly true of that species/strain.
3. EsViritu chooses exemplar sequences from multiple closely related genomes. The host field is not considered in this clustering/exemplar step. So the stated host for a sequence may not be representative for that cluster.

Nevertheless, I'm attaching a modified version of DB v3.2.4 with a "Host" field added. This field comes from the NCBI Virus resource. Please use caution in your interpretation.

virus_pathogen_database.host_metadata.v3.2.4.tsv.zip

[ZihengWu]

Thank you very much for the detailed and prompt reply @mtisza1 .

That makes a lot of sense, and I fully understand the limitations and potential biases associated with host information derived from public databases. This clarification is very helpful.

Just for reference, I wanted to share a few databases that curate virus–host associations, in case they might be of any interest in the future:

1. Virus–Host DB: https://www.genome.jp/virushostdb/
2. VirJenDB: https://virjendb.org/
3. gcPathogen: https://nmdc.cn/gcpathogen/

Thanks again for your time and for providing the host-annotated version of the database. I really appreciate it.

Best,
ZihengWu