kallisto task for RNA-seq

This task provides a convenience wrapper around kallisto. It can compute gene quantifications from reads in FASTQ format or build a kallisto index from transcript sequences in FASTA format.

Running

We encourage running this task as a Docker image, which is publicly available through GitHub packages. To pull the image, first install Docker, then run

docker pull docker.pkg.github.com/krews-community/rnaseq-kallisto-task/rnaseq-kallisto:latest

The task requires a pre-built kallisto index to be provided. A kallisto index can be built from one or more FASTA files with gene or transcript sequences by running this task:

docker run \
    --volume /path/to/inputs:/input \
    --volume /path/to/outputs:/output \
    docker.pkg.github.com/krews-community/rnaseq-kallisto-task/rnaseq-kallisto:latest \
    java -jar /app/kallisto.jar index \
        --sequence /input/seq1.fa --sequence /input/seq2.fa \
        --output /output/index.idx

Then, to perform quantifications, simply run:

docker run \
    --volume /path/to/inputs:/input \
    --volume /path/to/outputs:/output \
    docker.pkg.github.com/krews-community/rnaseq-kallisto-task/rnaseq-kallisto:latest \
    java -jar /app/kallisto.jar quant \
        --r1 /input/r1.fastq.gz --r2 /input/r2.fastq.gz \
        --index /input/index.idx --output-directory /output

This will produce an output directory containing quantifications (output.abundance.tsv).

Parameters

Indexing:

name	description	default
sequence	path to input FASTA sequences; many can be provided	(required)
output	path to output file; if it exists, it will be overwritten	(required)
kmer-size	the kmer size to use	31
make-unique	if set, sequences with duplicate names are made unique	false

Quantification:

name	description	default
r1	path to single-end or pair 1 reads in FASTQ format	(required)
r2	path to pair 2 reads in FASTQ format	(none)
output-directory	directory in which to write output	(required)
strandedness	set if reads are from one strand; forward, reverse, or unstranded	unstranded
output-prefix	prefix to use when naming output files	output
fragment-length	fragment length, required for single-end reads	(none)
sd-fragment-length	SD of fragment length, required for single-end reads	(none)
cores	number of cores available to the task	1
ram-gb	gigabytes of RAM available to the task	16

For developers

The task provides integrated unit testing, which quantifies a small number of reads against a human mitochondrial index and checks that the outputs match expected values. To run the tests, first install Docker and Java, then clone this repo, then run

scripts/test.sh

Contributions to the code are welcome via pull request.