frakka

Filter Kraken output files and calculate read-level and summary confidence score metrics per classified species.

Setting an appropriate Kraken2 confidence score threshold for species-level classification may appear rather arbitrary and will affect detection sensitivity. To take the guess work out of this, frakka creates per-species confidence score distribution metrics and plots.

Installation

GitHub

git clone https://github.com/stroehleina/frakka

Conda

TODO conda create -n frakka -c bioconda frakka

Usage

python frakka.py --help

usage: frakka.py [-h] [--version] [--kreport KREPORT] [--kout KOUT] [--species SPECIES] [--fof FOF] [--score SCORE] [--counts] [--taxid] [--plot] [--directory DIRECTORY] [--prefix PREFIX]
                 [--delim DELIM]

frakka - a tool to filter Kraken output files and calculate read-level and summary confidence score metrics per classified species

optional arguments:
  -h, --help            show this help message and exit
  --version, -v         show program's version number and exit
  --kreport KREPORT, -k KREPORT
                        Kraken2 report file(s) (comma-separated if multiple, required) (default: None)
  --kout KOUT, -o KOUT  Kraken2 output file(s) (comma-separated if multiple, required, must be in same order as --kreport / -k) (default: None)
  --species SPECIES, -sp SPECIES
                        List of corresponding true / known species names or taxids (comma-separated if multiple, optional, must be in same order as -k and -o files, single species provided assumes
                        all files are same true / known species) (default: None)
  --fof FOF, -f FOF     Tab-separated file of one (-k, -o, -sp)-tuple per line (default: None)
  --score SCORE, -s SCORE
                        Confidence score threshold, only reads / counts higher than this score are reported (default: 0)
  --counts, -c          Report total counts per species with score > --score / -s instead of per-read reporting (default: False)
  --taxid, -t           Species input and output are NCBI taxIDs instead of species names (default: False)
  --plot, -p            NOT IMPLEMENTED Plot distribution of score per species (default: False)
  --directory DIRECTORY, -d DIRECTORY
                        NOT IMPLEMENTED Specify output directory (default: .)
  --prefix PREFIX, -x PREFIX
                        NOT IMPLEMENTED Specify output file prefix (default: .)
  --delim DELIM, -del DELIM
                        Specify output file delimiter (default: )

Output

Tab-separated output to STDOUT:

file	read_id	true_spec_taxid	K_spec_taxid	score
file_1	read_1	species_A	species_A	0.6
file_1	read_2	species_A	species_B	0.4
file_1	read_3	species_A	species_A	0.6
file_2	read_a	species_C	species_C	0.6
...	...	...	...
file_n	read_m	species_X	species_Y	0.01

Tab-separated output to STDOUT, when --counts (and optionally, --score) are specified:

file	K_spec_taxid	read_count (of reads above --score)	median_score (of reads above --score)
file_1	species_A	count_1A	0.6
file_1	species_B	count_1B	0.4
file_1	species_C	count_1C	0.6
file_2	species_A	count_2A	0.6
...	...	...	...
file_n	species_X	count_nX	0.01

Logfile goes to STDERR.

Graphical output

TODO

Plot distribution of score per species.

Shiny confidence score sweep

TODO

Shiny app to allow to adjust confidence score and see how this affects number of reads binned into (selectable set of) species.

Etymology

Frakka is partly "filter Kraken", partly Old Norse for "spear" and partly inspired by the Torstyverse suite of tools by Torsten Seemann which follow a similar naming pattern.