A post-processing tool for Kraken 2 read classifications. Run Kraken 2 once and re-classify the reads with any confidence score stringency of your choice afterwards, saving you lots of compute time. Creates Kraken 2 style report and read classification files.
For additional insight into your Kraken 2 classifications, try out KrakMeOpen - a downstream analysis toolkit for Kraken 2 classification quality metrics.
StringMeUp is available to install through conda. Simply run the following command to install it:
conda install -c conda-forge -c bioconda stringmeup
A good start is to run stringmeup --help.
The confidence score (CS) for a given read R classified to a given node J is calculated by dividing the number of k-mers that hit any node in the clade rooted at node J (N) by the total number of k-mers that were queried against the database (M). Any k-mer with an ambiguous nucleotide is not queried against the database, and is thus not part of M.
CS = N / M
If the CS for a given read R at a given node J is equal to or larger than the specified cutoff, read R is classified to node J. If not, the CS of read R is calculated for the parent of node J. This is repeated until the CS >= CS cutoff or until we reach the root of the taxonomy. If the CS < CS cutoff at the root, the read is deemed unclassified.
To reclassify reads classified by Kraken 2 with a confidence cutoff of 0.1:
stringmeup --names <names.dmp> --nodes <nodes.dmp> 0.1 <original_classifications.kraken2>
Where:
- original_classifications.kraken2 is the output file from Kraken 2 that contain the read-by-read classifications.
- names.dmp and nodes.dmp are the same NCBI taxonomy files used for the building of the database that was used to produce the classifications in original_classifications.kraken2.
This command would output a Kraken 2 style report to stdout. Adding --output_report <FILE> would save the report in a file.
To save the read-by-read classifications, add --output_classifications <FILE> to the command.
To save a verbose version of the read-by-read classifications, add --output_verbose <FILE> to the command. The verbose version of the read-by-read classifications will contain the following columns:
| Column | Explanation |
|---|---|
| READ_ID | The ID of the read |
| READ_LENGTH | The length of the read (same as Kraken 2 output) |
| MINIMIZER_HIT_GROUPS* | The number of minimizer hit groups found during Kraken 2 classification* |
| TAX_LVL_MOVES | How many levels in the taxonomy that the read moved during reclassification |
| ORIGINAL_TAXID | The taxID that the read was classified to originally |
| NEW_TAXID | The taxID that the read was reclassified to |
| ORIGINAL_CONFIDENCE | The original confidence score |
| NEW_CONFIDENCE | The confidence score at the taxID that the read was reclassified to |
| MAX_CONFIDENCE | The maximum confidence that the read can have |
| ORIGINAL_TAX_LVL | The taxonomic rank of the orignally classified taxID |
| NEW_TAX_LVL | The taxonomic rank of the reclassified taxID |
| ORIGINAL_NAME | The scientific name of the original taxID |
| NEW_NAME | The scientific name of the reclassified taxID |
| KMER_STRING | The k-mer string (same as Kraken 2 output) |
*: Is only present if the forked version of Kraken 2 was used for initial classification.
This option requires an input file that was produced with my fork of Kraken 2.
Add --minimum_hit_groups <INT> to the command. A read can only be considered classified if the number of minimizer hit groups is at or above the minimum_hit_groups setting.