search

epi2me-labs / wf-human-variation

Other
86 stars 41 forks source link

Does the phasing run despite the configuration not to run phasing? #160

Closed ilivyatan closed 2 months ago

ilivyatan commented 3 months ago

Ask away!

Hi, In my experience with the earlier versions of the workflow, the phasing stage takes a considerable amount of time and dramatically slows the workflow ( days versus hours of runtime). Therefore, I run it with all the phasing settings off, and then run phasing post-hoc to the workflow. I've updated to the latest workflow version (11) and the behavior of the long run time has returned. I reached out to the support and they informed me that running with --str for repeat expansion analysis, will initiate a phasing step even if I haven't configured one. So now, I've only run with --sv --snv --cnv. But the runtime is still very long (Over 20 hours for a ~20x covered human genome on a cloud platform with 16 cores, 64GB RAM, 600 GB storage). How can I run the workflow now without invoking any phasing? Thanks!

RenzoTale88 commented 3 months ago

Hi @ilivyatan running the workflow with the --snp --sv --cnv options should not trigger the generation of phased genotypes and haplotagged reads, unless explicitly requested with --phased.

However, please notice that there is still a phasing step as part of the Clair3 software, which cannot be skipped as it is an intrinsic part of the Clair3 software.

The processes involved in the generation of phased variants are the following: 1) post_clair_phase_contig 2) post_clair_contig_haplotag 3) phase_all

Do you see any of them running?

ilivyatan commented 3 months ago

Actually, I took a closer look and it's the cnv_spectre:callCNV (0 of 1) that just stays at 0% for over 16 hours so far. I ran it again with only --snv and --sv and it finished after 1.5 hours. Why isn't the CNV analysis advancing?

RenzoTale88 commented 2 months ago

This is a distinct issue, please create a new ticket providing the information requested.