Lightning

ShellCommandRequirement enables cwl to parse shell characters like && or |, see http://www.commonwl.org/v1.0/CommandLineTool.html#ShellCommandRequirement.

Review of su92l-xvhdp-zb7seb8e430gfmy beagle.cwl:

• Do commands look right/use appropriate flags and resources?
  • Yes ✓ It's just one beagle command && tabix command to output indices. The beagle jar file is in the docker container.
  • 1 thread, default window size ✓
  • 1k genomes reference in bref format ✓
  • Genetic map is GRCh37, which is fine for all but chrM for HG19 genome ✓
  • Target is vcf of chromosome 19 of GS12877, which has HG19 for reference. ✓
  • Dockerfile in l7g-ml repo looks fine. ✓
• Do logs show any errors?
  • None ✓
• Does output look correct? (Cannot "grade" accuracy)
  • Desired outputs are captured. Imputed VCF looks correct. stdout for beagle is just some information about time on each phasing window, so sending to log is fine. (i.e. info about timing could be retrieved from log)
• Is output meaningful?
  • Input has already been phased by eagle, so presumably beagle performs no phasing (unless beagle is phasing a position that was passed through by eagle, or overwriting the phasing of eagle; but this is unlikely because eagle & beagle are using the same reference panel). ✓
  • Output is indeed phased and imputed. ✓
• Open-ended: How were errors I encountered overcome?
  • Main problem encountered was ERROR:duplicate allele at 19:_____, but that was for an unphased input. Perhaps eagle discards them?
• Remaining issues or questions:
  • 1k genomes data is technically whole genome, but in VCF, beagle (and perhaps eagle as well), ignores END= tags (runs of reference).