how does kraken differentiate masked genomes

[zhaoc1]

Hi,

I got the masked genomes from the eupathDBclean, and built a kraken database on it (kmer = 25). By looking back at the reports, some of the kmers fully mapped to the masked sequences, e.g.

AFPV01005151 |
NNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNN

Therefore, I am wondering how does Kraken deal with the masked (low complexity) genome regions? How can I filter out those false positive kmer results that mapped to the masked region? Thank you!

Best,
Chunyu

[jenniferlu717]

That shouldn't happen. Any kmers with "N" don't end up in the database.

[zhaoc1]

I actually build a krakenHLL database on the cleaned eupathDB and this kmer mapping show up in my reports. Do you know why this is happening? Thanks!

[jenniferlu717]

Can you send me both the read causing this and the Kraken output showing this match? Please email the Kraken line and the read to jennifer.717@gmail.com

[fbreitwieser]

@zhaoc1 , maybe this is a KrakenHLL issue. You made a database with nodes for sequences and genomes, right? Did you restart the building process? Maybe the mapping file is not up to date.

[zhaoc1]

@fbreitwieser this is how I built the KrakenHLL database, after downloading the eupathDBclean files and seqid2taxid.map

DBNAME=testDB
krakenhll-download --db $DBNAME taxonomy
cp my_eupathDB_folder/*. $DBNAME/library
cp seqid2taxid.map $DBNAME/library
krakenhll-build --db $DBNAME --taxids-for-genomes --taxids-for-sequences --kmer-len 25 --threads 8

Could you please explain during which step the mapping file is not up to date, or is there any potential problems here? Thank you!