Sharing Refseq Virus taxonomy/metadata file?

[haleyhallowell]

Hello! I have been really enjoying sylph so far, thank you for a great contribution to the field!

In the paper, i saw that you used the Refseq viral database. I am interested in using this as well, and was wondering if you'd be able to share your taxonomy/metadata file for the mpa format processing script? I have the refseq viral database built for sylph, but thought i would ask before i went through the process of building the metadata file! Thanks!

[bluenote-1577]

Hi @haleyhallowell

I actually never built the Refseq viral taxonomy metadata file. My results for the paper were all in the default tsv output (since I just looked at the total abundance of viruses, not specific taxa).

If you have a viral metadata file/database, you're welcome to share it as well.

Be careful with how viruses are processed though, in https://github.com/bluenote-1577/sylph-utils/blob/main/sylph_to_taxprof.py there is some pre-processing of viral genome names to ensure compatibility with IMG/VR. You may have to modify the script slightly to make it compatible with your format. Letm eknow if you have questions

[haleyhallowell]

Thanks so much for the reply! if i can successfully build the metadata file ill be happy to share! I took a look at the suggestion you had for building the metadata from refseq and it looks a little above my skill level. I'm thinking of parsing the species/genome level names out of the fasta headers of the .fna file. do you think something like this in a .tsv file would work? The reason i am interested in using the utility script is the ability to reformat the data and then combine multiple files together (similar to merge_metaphlan_tables.py) so the actual full taxonomy does not matter that much to me

Example header:

NC_001798.2_Human_herpesvirus_2_strain_HG52,_complete_genome

[bluenote-1577]

If you do not care that much about the actual taxonomy, then you can just analyze sylph's output tsv file -- look at the last column "Contig_name which should be NC_001798....

You can run sylph profile database.syldb reads1.fq.sylsp reads2.fq.sylsp .... -o all_results.tsv. So all_results.tsv has all profiles mixed together.

Then you can use R/python to manipulate all_results.tsv by analyzing the Sample_file and Contig_name columns. If you want like a matrix of results with TAXA x SAMPLE then you can use python/R to manipulate the all_results.tsv file. Ask ChatGPT , for example :)