kracken2/bracken output file integrity check

[Sunflower34]

Hi everyone,

I just ran Kraken2 and Bracken on a large metagenomics dataset on a SLURM cluster, and since some jobs might have suffered from silent memory crashes or timeouts, I'd like to make sure none of the output files were truncated or corrupted due to memory issues, timeouts, or failed jobs.

Are there any recommended tools or workflows for checking the integrity and completeness of Kraken2/Bracken reports ?

Thanks for your help!

[williamcorlin]

One simple thing I do is check the SLURM logs first. If the job finished normally and there are no OOM (out-of-memory) or timeout messages, that's already a good sign.

For Kraken2 reports, I also verify that:

• The report ends cleanly (not cut off mid-line).
• The total percentage is close to 100%.
• The total classified + unclassified reads matches the expected number of input reads.

car games
For Bracken outputs, compare the total estimated reads against the Kraken2 counts and look for unusually small files or samples with suspiciously low counts compared to the rest of the batch.

If you're processing many samples, a quick script that checks file size, line count, and expected report format can help identify outliers. In my experience, truncated files usually stand out immediately because the report structure is incomplete or the totals don't make sense.