Question about assembly level in creating flextaxd database

[ingridvanw]

Hi guys,

Thank you for developing this tool.

I have one question. I downloaded genomes with all types of assembly from NBCI (Refseq).
When created the flextaxd database, I noticed in the inspect.txt file that genomes with assembly level contig are not in there.
It is possible to create a flextaxd database with genomes with assembly-level 'contig'? (Of is the 'minimum level' scaffold?)

Commandlines I have used
datasets download genome taxon --assembly-source RefSeq --assembly-level chromosome,complete,scaffold,contig Mycolicibacillus --filename mycolicibacillus.zip
--> download 7 genomes

flextaxd --db mycolicibacillus.fdb --taxonomy_file nodes.dmp --taxonomy_type NCBI --genomeid2taxid nucl_gb.accession2taxid.gz --genomes_path genomes

flextaxd --db mycolicibacillus.fdb --clean_database --taxonomy_type NCBI

flextaxd-create --db mycolicibacillus.fdb --genomes_path genomes --dbprogram kraken2 --create_db --db_name kraken2.mycolicibacillus --processes 20
--> mycolicibacillus trivialis is not in inspect.txt (GCF_002102395.1)

Thanks in advance!
Kind regards
Ingrid

[ingridvanw]

@davv2
I hope you can comment on this :)?
Thanks in advance!

[davve2]

Hi Ingrid,

sorry for a delayed response. Flextaxd is not really doing any selection on assembly level, what may cause the issue is that when annotating the genomes with nucl_gb it doesn´t include the correct annotation for the sequence in question. To ensure all sequences being annotated from NCBI I recommend to download and merge nucl_gb.accession2taxid.gz with nucl_wgs.accession2taxid.gz

You can find a description on how to do that here https://github.com/FOI-Bioinformatics/flextaxd/wiki/8d.-Walkthrough--merge-NCBI-and-GTDB under section 1.2

Let me know if this is not solving your problem and I will try to look into it.

Kind regards,
David

[DariaOGajewska]

Hi @davve2,

I encountered a similar problem. I have a genome dataset of about 20k sequences and after "cleaning", 4k sequences of it were removed. After that kraken2 database was created from a flaxtaxd file the database weighs about 50GB.

Then I tried David's method to concatenate the "nucl_gb.accession2taxid" and "nucl_wgs.accession2taxid". I build the database with "flextaxd-create" exactly the same as before but with a new accession2taxid, I drastically ran out of storage space, and it was around 700GB.

Is this extreme increase in storage usage expected during the database-building process with a concatenated accession2taxid file?

Any insights or suggestions would be greatly appreciated!

Kind regards,
Daria

[davve2]

Hi @DariaOGajewska @ingridvanw

This is expected functionality. The number of sequences supplied is not correlating linearly to the database size. The database consists of unique kmers (or minmizers) which rather extend the database with additional "variation". The 4k sequences that weren't annotated in the first example are genomes that is considered not having the same quality as the genbank dataset, this could for example be incomplete genomes or meta assembled genomes.

What it looks like to the result is that this dataset contain a significant addition of variation which is not present in the genbank "filtered" dataset, perhaps it will include additional genus or species to the dataset. So this "problem" is not really a FlexTaxD or kraken2 issue but rather just the fact that the nucl_wgs dataset contains additional "genome" variation not present in the genbank dataset.