Medaka variant error: "IndexError: index 0 is out of bounds for axis 0 with size 0"

[BioWilko]

Describe the bug A clear and concise description of what the bug is including the command that you have run.

medaka variant crashes with exit status 20 when the command medaka variant MPXV.reference.fasta MPVX_untreated_LSK114_280722-fast_barcode74.1.hdf MPVX_untreated_LSK114_280722-fast_barcode74.1.vcf is run.

Logging

Traceback (most recent call last):
  File "/home/ubuntu/miniconda3/envs/artic-test/bin/medaka", line 11, in <module>
    sys.exit(main())
  File "/home/ubuntu/miniconda3/envs/artic-test/lib/python3.8/site-packages/medaka/medaka.py", line 720, in main
    args.func(args)
  File "/home/ubuntu/miniconda3/envs/artic-test/lib/python3.8/site-packages/medaka/variant.py", line 230, in variants_from_hdf
    for sample in joined_samples:
  File "/home/ubuntu/miniconda3/envs/artic-test/lib/python3.8/site-packages/medaka/variant.py", line 104, in join_samples
    yield medaka.common.Sample.from_samples(queue + [to_yield])
  File "/home/ubuntu/miniconda3/envs/artic-test/lib/python3.8/site-packages/medaka/common.py", line 143, in from_samples
    rel = Sample.relative_position(s1, s2)
  File "/home/ubuntu/miniconda3/envs/artic-test/lib/python3.8/site-packages/medaka/common.py", line 224, in relative_position
    s1_ord, s2_ord = sorted((s1, s2), key=lambda x: (x.first_pos, -x.size))
  File "/home/ubuntu/miniconda3/envs/artic-test/lib/python3.8/site-packages/medaka/common.py", line 224, in <lambda>
    s1_ord, s2_ord = sorted((s1, s2), key=lambda x: (x.first_pos, -x.size))
  File "/home/ubuntu/miniconda3/envs/artic-test/lib/python3.8/site-packages/medaka/common.py", line 89, in first_pos
    return self._get_pos(0)
  File "/home/ubuntu/miniconda3/envs/artic-test/lib/python3.8/site-packages/medaka/common.py", line 69, in _get_pos
    return p['major'][index], p['minor'][index]
IndexError: index 0 is out of bounds for axis 0 with size 0

Environment (if you do not have a GPU, write No GPU):

• Installation method: conda (mamba solver)
• OS: Ubuntu 20.04 LTS
• medaka version: 1.6.1
• GPU model: Nvidia A100
• Nvidia driver version: 510.73.05
• CUDA version: 11.6
• cuDNN version: 8.4.1.50

Additional context This problem only seems to have appeared after changing MPX reference from AY753185.1 to MT903344.1 which included translating the primer bed file to have co-ordinates relative to the new reference (potentially affecting primer trimming). This issue only occurs for particular barcodes on particular runs. I am happy to share data privately to help locate the issue but cannot share the data publicly.

[SamStudio8]

Hi @BioWilko, fancy seeing you here! I believe I have narrowed down the scope of this issue to a slicing operation in join_samples. I have crafted a dummy HDF5 file that can trigger the same IndexError you have provided. It would be good to understand how this could happen on a real dataset so I have sent a request to the e-mail address on your Github profile which should contain instructions to provide your HDF to us.

[BioWilko]

Small world eh @SamStudio8? Predictably the error refused to manifest when I tried to replicate it but I managed it in the end and have sent over the file, thanks!

[SamStudio8]

Your data triggered an interesting edge case wherein the Sample chunk generator was emitting a chunk containing only a long insertion. Compared against the reference the chunk had only one "major" non-variant position at index 0. When the Sample was bisected on index 0, an empty Sample slice was created (with no indexable positions) and you know the rest.

I have patched Medaka to explicitly avoid this case and released v1.7.1 (which you can get from the releases page, or via bioconda). Please advise if this has fixed your issue.

[BioWilko]

Sorry I took so long to get back to you but I have finally tested this and confirm it has fixed my issue, thanks!