How to Build a Customised Kraken2 Database

M. Asaduzzaman Prodhan^*

Step 1: Download the genomes

👉 How to Retrieve Public Data

Keep these genome files (.fna or .fa) in a single directory.

Step 2: Download the taxonomy

Kraken2 requires taxonomy files from NCBI to classify sequences.

Below is a Slurm batch script for downloading taxonomy files on an HPC cluster:

```

!/bin/bash --login

SBATCH --job-name=K2DB # Name of the job

SBATCH --account=XXXX # Project account

SBATCH --partition=work # Partition (queue) to run the job

SBATCH --time=1-00:00:00 # Time limit (1 day)

SBATCH --ntasks=1 # Single task/job

SBATCH --ntasks-per-node=1

SBATCH --nodes=1 # Request 1 node

SBATCH --exclusive # Request exclusive access to the entire node (all CPUs), alternative --cpus-per-task=128

SBATCH --no-requeue

SBATCH --export=none

Activate the Kraken2 Conda environment

conda activate kraken2

Download the taxonomy file

Define DB directory

DBDIR=kraken2DB export KRAKENDBNAME=$DBDIR # <-- important fix

Step 1: Create DB directory and download taxonomy

kraken2-build --db $DBDIR --download-taxonomy --threads 120 --use-ftp

End of script

```

Explanation:

• conda activate kraken2 → loads Kraken2 environment

• export KRAKENDBNAME=$DBDIR → tells Kraken2 where to store the database

• kraken2-build --download-taxonomy → fetches NCBI taxonomy files

• --threads 120 → uses 120 CPU threads for faster download

• --use-ftp → ensures NCBI files are retrieved via FTP

Step 3: Add the genomes to library

Once you have genomes downloaded, you need to add them to your Kraken2 database.

```

!/bin/bash --login

SBATCH --job-name=K2DB

SBATCH --account=XXXX

SBATCH --partition=work

SBATCH --time=1-00:00:00

SBATCH --ntasks=1

SBATCH --nodes=1

SBATCH --exclusive

SBATCH --no-requeue

SBATCH --export=none

Activate Kraken2 environment

conda activate kraken2

Fix Kraken2 environment variables

DBDIR=kraken2DB export KRAKENDBNAME=$DBDIR export KRAKEN_DIR=/path/to/miniconda3/envs/kraken2/libexec

Add the genome(s) to the library

find . -name "*.fna" -print0 | \ xargs -0 -I {} kraken2-build --db $DBDIR --add-to-library "{}" --threads 120 ```

Explanation:

• find . -name "*.fna" → looks for all .fna files in current directory

• xargs ... kraken2-build --add-to-library → adds each genome file into the Kraken2 library

• --threads 120 → speeds up processing when handling many genomes

Step 4: Generate the seqid2taxid.map file

Kraken2 needs a mapping file between genome sequence IDs and taxonomy IDs. This ensures correct classification.

Step 4.1

wget ftp://ftp.ncbi.nlm.nih.gov/pub/taxonomy/accession2taxid/nucl_gb.accession2taxid.gz

Explanation:

• NCBI provides accession2taxid files mapping every accession to its taxonomy ID

Step 4.2

gunzip nucl_gb.accession2taxid.gz

Step 4.3

cut -f2,3 nucl_gb.accession2taxid > seqid2taxid.map

Explanation:

• Kraken2 needs a slimmed-down seqid2taxid.map

• Only column 2 (accession.version) and column 3 (taxid) are kept

Step 4.4

Copy the seqid2taxid.map to $Kraken2DB (not in the taxonomy or library directory)

Step 5: Build the DB

Now that taxonomy and genome libraries are in place, the final step is to compile the database.

```

!/bin/bash --login

SBATCH --job-name=K2DB # Name of the job

SBATCH --account=xxxx # Pawsey project account

SBATCH --partition=highmem # Partition (queue) to run the job

SBATCH --time=1-00:00:00 # Time limit (1 day)

SBATCH --ntasks=1 # Single task/job

SBATCH --ntasks-per-node=1

SBATCH --nodes=1 # Request 1 node

SBATCH --exclusive # Request exclusive access to the entire node (all CPUs), alternative --cpus-per-task=128

SBATCH --no-requeue

SBATCH --export=none

Activate Kraken2 environment

conda activate kraken2

Fix Kraken2 environment variables

DBDIR=kraken2DB export KRAKENDBNAME=$DBDIR export KRAKEN_DIR=/path/to/miniconda3/envs/kraken2/libexec

Build the database

kraken2-build --db $DBDIR --build --threads 120

End of script

```

Explanation:

• kraken2-build --build → processes taxonomy + genomes into a searchable DB

• --threads 120 → uses multiple cores for faster DB building

• DBDIR naming convention (e.g., kraken2DB) helps version-control DBs

At this stage, your custom Kraken2 database is ready for classification!