Investigate HTPC friendly method for parallelisation

[timothymillar]

Current multi-core method requires a single master python process to start sub-processes. This means that the entire job needs to be submitted as a single chunk of HTPC resource. It would be good to have an option to run in a way that is more friendly for splitting into many jobs with a tool like asub.

The best way to achieve this is with an approach similar to freebayes. Things to work out

• [x] specify a single region instead a whole bed file
  • specify a region string and optional id ala freebayes?
  • pass a bedfile line interpreted as an argument of 3-4 values?
• [x] option to specify a sample-bam map to minimise io
  • should conflict with bam/sample arguments
  • avoids need to check bam headers for samples present in each process
  • order within a sample-bam map file should dictate output sample order
• [ ] include an example or script for using asub

Output per process would be VCF with a single record, these can be merged with vcflib

[timothymillar]

Example usage:

# array job
for TARGET in $(cat targets.bed)
do
# unique name for VCF containing each record
VCF="${TARGET//'\t'/_}.vcf.gz"
cat << EOF
mchap assemble \
    --target "$TARGET" \
    --variants "variants.vcf.gz" \
    --reference "reference.fasta" \
    --sample-bams "sample_bams.txt" \
    --sample-ploidy "sample_ploidy.txt" \
    --best-genotype | bgzip > "$TMPDIR/$VCF"
EOF
done | asub -c 60 -j mchap-asub

[timothymillar]

Use of sample-bam map file also prompts the requirement for a sub-tool that can create such a file from a list of bam files and an optional list of samples.

[timothymillar]

Done in #100