Investigate regions of interest on chr6.

[samuelklee]

MHC, LPA, etc.

HPRC+AoU1 chr6 KAGE panel and the input panel VCF can be found in this submission: https://app.terra.bio/#workspaces/allofus-drc-wgs-lr-prod/AoU_DRC_WGS_LongReads_PacBio%20PAPER%20COPY/job_history/d9d68713-3c80-49f6-a68b-c8fc2afae971
(note related runs https://app.terra.bio/#workspaces/allofus-drc-wgs-lr-prod/AoU_DRC_WGS_LongReads_PacBio%20PAPER%20COPY/job_history/dea0b0fd-fa39-400e-8a10-655be8c99eb2 and https://app.terra.bio/#workspaces/allofus-drc-wgs-lr-prod/AoU_DRC_WGS_LongReads_PacBio%20PAPER%20COPY/job_history/dad56157-90fd-4862-8f92-0c43fd8fdd0d; the former successfully created the panel VCF, but failed due to undersized evaluation VMs, which were corrected in the successful latter run).

[samuelklee]

Some notes to self:

In prepping for imputing this chr6 panel into 1kG in the new imputation workspace (https://app.terra.bio/#workspaces/allofus-drc-wgs-LR-prodPaper/AoU_DRC_WGS_LongReads_Imputation), I updated the evaluation WDLs to use the more streamlined case workflow introduced in https://github.com/broadinstitute/kage-lite-development/pull/29/files; previously, those WDLs used a slightly more outdated and less efficient version that did not utilize named pipes. But, a few caveats:

1. I believe the new version only drops reads with N nucleotides, rather than read pairs, which means things don't tie out exactly.
2. You must be sure to use a KAGE Docker that has the correct version of samtools, otherwise the task with named pipes will fail silently (the possibility of which I noted in Updated samtools in Docker and optimized I/O and pipelining in case WDL. kage-lite-development#29 (comment)). Probably we need to do something along the lines of https://stackoverflow.com/questions/53450561/capture-non-zero-exit-code-from-named-pipe.

Despite minor differences, performance still looks good, so I will kick off 50 trios today.