Blastn for all bacteria genomes; create the db

Question

[davidmaimoun]

Hello, I need to run blastn on Bacteria to see if my sample really is nesseirai men, and didn't have been contaminated. I downloaded reference seq from ncbi for each bacteria. In a data folder, got many subfolders, each one contains a fasta file (.file), representing a genome.

subfolders

genome file

How can I build a db with this data? I tried the cmd makeblastdb for each genome but it doesn't seem to me the good way.

Thank you

Answer 1

[GenoMax]

I tried the cmd makeblastdb for each genome but it doesn't seem to me the good way.

If you have made the individual databases then use blastdb_alias tool to create a common alias : https://www.ncbi.nlm.nih.gov/books/NBK569848/

Otherwise walk through the folder structure, cat the .fna files into one large fasta and then create one database.

[davidmaimoun]

Thank you GenoMax !

After running makeblastdb have got this:

after the cmd

in each folder (~3500 folders).

Do you advise me to use the alias? Or it's better to create a big multi-fasta and create the db based on it?

[GenoMax]

If all of your databases have that identical name then you will need to rename/recreate them. At that point you may as well cat the .fna and re do the blastdb. If you used a script to make the 3500 db you could modify it so the output files gain unique names. You will have to tell us if the aliastool works for 3500 databases.

[davidmaimoun]

Thank you for the answers guys. I went with the cat .fna solution and it fit well to what I have to do. Again, thank you!

[GenoMax]

Please accept this answer (green check mark) to provide closure for this thread.

Answer 2

[shenwei356]

I'd recommend using some other tools:

• sourmash gather
• sylph

They can both compute "signature" from the reads and genomes and analyze the microbial composition. Note that, for querying with genomes, small thresholds should be used.

[GenoMax]

Based on a different thread these are not metagenomic samples. OP is interested in checking if there is any contamination besides one expected genome (from what I gather it is not normally expected). We already discussed that none of the tools (including blast) are likely to be good for that purpose, especially if the "contamination" is from a close relative.

[shenwei356]

see if my sample really is nesseirai men, and didn't have been contaminated

OP might have some assemblies of supposed-to-be single-specie data, and they might be mixed with contigs from other species. So we can treat it as metagenomic data, in contigs rather than reads, where a metagenomic profiling tool might help.

We already discussed that none of the tools (including blast) are likely to be good for that purpose, especially if the "contamination" is from a close relative.

Yes, close relatives are hard to tell.