Empty fields in Kraken2 report file generated with version 2.1.3 and PlusPF database break Bracken downstream analysis

[Seb-vb]

Hello,

I believe this issue is the same one as in #883. I hope to add some relevant details to make the problem clearer.

I am using Kraken2 version 2.1.3 with the PlusPF database(Sept 2024 version) and am encountering a problem with the generated report file (example below). This problem affects downstream analysis as Bracken is unable to process the report file without manual correction.

The fourth column is supposed to contain an abbreviation specifying the taxonomic level of each taxon. However some cells are empty, such as for "root" and "Bacteria". They should contain the letters "R" and "D" respectively. A few other cells, such as the one for "cellular organisms" in this example only contain a digit, when the value should be "R1". This problem occurs a few times in the whole report, consistently with Domains, and sporadically in other taxonomic levels

1	2	3	4	5	6
35.93	2548128	2548128	U	0	unclassified
64.07	4542966	6838		1	root
62.13	4405890	999	1	131567	cellular organisms
61.73	4377315	62068		2	Bacteria
28.7	2034982	33109	D1	1783272	Terrabacteria group
26.75	1896932	4877	P	1239	Bacillota
13.33	945053	377	C	909932	Negativicutes

As mentioned, Bracken crashes when using the incomplete report. It however works correctly after manual correction.

I encountered this problem when running the program on different samples from different sources so I'm fairly confident they aren't the cause. I'm not sure however if this problem comes from Kraken2 itself, the PlusPF database, or an error on my side. What are your thoughts ?

I hope this was clear enough, please tell me if any other details can be useful.

[Xiang-Leo]

I met the same problem when I used the standard databse. I tried to add the Domain mannuly but there are a lot of species which should be put different family or order, were put "D".

[Xiang-Leo]

After check, I confirmed that the newest version will generate this kind of result. Using conda to install version 2.0.8 will resolve all problems.

[Seb-vb]

I didn't think to check previous versions of Kraken2, thanks for the information @Xiang-Leo.

On my side I made a script to fix the 2.1.3 k2 report by assigning the correct taxonomic level when needed. It uses the "ktaxonomy.tsv" file that is included with the pre-built databases which very conveniently contains the abbreviated taxonomic levels of every taxon of the database.

The best workaround right now does seem to simply use version 2.0.8. Hopefully this error will be possible to fix for the next release.

[meyersbs]

I'm also experiencing this issue with kraken2 v2.1.3 and bracken v2.9

@Seb-vb Can you share your script for fixing the report?

[meyersbs]

I'm also running into this issue with kraken2 v2.0.8 and v2.1.2