search

iqbal-lab-org / gramtools

Genome inference from a population reference graph
MIT License
92 stars 15 forks source link

quasimap Assertion `index_end_boundary >= allele_coverage_offset' failed #121

Closed martinghunt closed 6 years ago

martinghunt commented 6 years ago

gramtools version b05f148e0ecd522e6d1dc1affbc3b8798320091f

commands:

source /nfs/leia/research/iqbal/mhunt/Cryptic_releases/201805/Multi_sample/python_venv/bin/activate

gramtools quasimap --gram-directory /hps/nobackup/iqbal/mhunt/tmp/multi_sample.nextflow.work/76/55ee77e59995f90e862352526d38f0/small_vars_clustered.gramtools.build/split.27.gramtools_build --output-directory test.out --reads small_vars.minos.132/split.out/unmapped_reads.bam --reads small_vars.minos.132/split.out/split.27.reads.bam

Ran in here: /hps/nobackup/iqbal/mhunt/tmp/multi_sample.nextflow.work/ca/afe74136db866b6432211db8db35f1/

See test.sh.{o,e} in that directory for stdout/err and LSF info

ffranr commented 6 years ago

I can reproduce this locally (my laptop). Will debug.

ffranr commented 6 years ago

I have a partial solution for this issue. I understand it completely. However, we need to readdress https://github.com/iqbal-lab-org/gramtools/issues/90 before I can proceed with certainty. I've left a comment on that issue to further enhance the spec.

ffranr commented 6 years ago

@martinghunt @iqbal-lab This bug arises when a single read has multiple mappings which are encapsulated within different alleles.

Gramtools attempts to allocate PRG paths to mappings which are encapsulated within alleles. It failed to associate read mapping SA indexes with the correct alleles.

The solution has been merged into master here: https://github.com/iqbal-lab-org/gramtools/commit/e22cd4fbddadbbbb05ecc5bbe4fc91880eb128f5

Will tag this issue with testing and keep open for a week or two or until feedback is obtained.

martinghunt commented 6 years ago

Update: running on the same dataset that caused the error. All samples so far (~150 of 4000) ran ok, including the one above that had the error.

martinghunt commented 6 years ago

I'm sure it's fine. 4016/4034 ran OK. The nextflow job that was orchestrating all these gramtools runs got killed by LSF because it hit the run time limit before the remaining 18 could finish.