Daisy Documentation¶
Daisy is a framework to perform computational experiments efficiently, reproducibly, and at scale. An experiment is defined by an experimental design in yaml format that describes one or more tools to be run on one or more data sets and collecting on or more metrics from the results.
At its simplest, an experimental design would look like this:
setup:
tools:
- freebayes
- octopus
metrics:
- bcftools_stats
input:
bam: *.bam
reference_fasta: hs37d5.fa
This would apply the variant callers freebayes and octopus to all files ending in .bam in the current directory and running the tool bctools stats on the results. This is a typical benchmarking scenario comparing to variant callers.
The same experimental design can be used for parameter optimization, for example:
setup:
tools:
- freebayes
metrics:
- bcftools_stats
input:
bam: *.bam
reference_fasta: hs37d5.fa
freebayes:
options:
- --haplotype-length 50
- --haplotype-length 100
This design would run freebayes with two different options.
While originally developed for benchmarking, we have found the same framework useful in processing large data sets. For example, to map a multiple read data sets from a high-throughput sequencing experiment, you could say:
setup:
tools:
- bwa_mem
metrics:
- samtools_stats
input:
fastq: *.fastq.gz
reference_fasta: hs37d5.fa
group_regex: /([^/]+)-R.*.fastq.gz
group_alias: \1
The regular expression ensures that the two components of a sample (named <sample>-R1.fastq.gz and <sample>-R2.fastq.gz are grouped and supplied together to the mapping tool bwa mem. The samtools stats command is run to provide QC metrics such as the proportion of reads mapped to the reference genome.