bracken-build: ...cat: write error: Broken pipe

[dinhe878]

Hi,

I tried several times with the command bracken-build -d . in my krakenDB folder and it always ended up with the error message:

Selected Options: kmer length = 35 read length = 100 database = . threads = 20 Checking for Valid Options... Creating database.kraken [if not found] >> kraken2 --db . --threads 20 <( find -L ./library ( -name *.fna -o -name *.fa -o -name *.fasta ) -exec cat {} + ) > ./database.kraken Loading database information... done. Processed 181624 sequences (3959379151 bp) ...cat: write error: Broken pipe

According the the manual page, it seems to be the step 1a somehow hanged for a long time and eventually got killed. I have used the brackeen-build previously on a standard kraken2 DB and it was fine. So I suspected that it has something to do with this custom-build kraken DB which was built using kraken2-build with kraken-predefined refseq archaea, bacteria, viral, human, fungi, plant, protozoa, and additional 48 insect genomes. The only difference in the /library content that I can find between the standard kraken DB and the custom kraken DB was that there are more than one .fna files in the additional /library/added folder, whereas the standard archaea, bacteria, viral, human, fungi, plant, protozoa all only have one library.fna file.

I am not sure what was the problem causing this error. Need help! Thanks!

[dinhe878]

So I ended up just search all my fna files against my custom krakenDB separately and concatenate them, like

kraken2 --threads=40 --db=. ./library/viral/*.fna > database.kraken kraken2 --threads=40 --db=. ./library/added/*.fna >> database.kraken

And it turned out that the error might come from the plant sequences, as

kraken2 --threads=40 --db=. ./library/plant/*.fna >> database.kraken Loading database information... done. Processed 165820 sequences (1865041149 bp) ...Killed

I don't really mind losing plant taxonomy classification in this case, just went ahead proceed the rest of the build and it worked. But the error remains mysterious...

[jenniferlu717]

@dinhe878 apologies for not responding earlier, I've been preoccupied with other projects.

I'll look further into this issue but thank you for letting me know