Hisat2 align on large index

[andreott]

Snakemake version
snakemake-wrappers: 3.3.3

Describe the bug
The wrapper looks for the index file checking for .ht2 file extension only. So this fails when a large index was produced with extension .ht2l

Fixing
A quick and dirty fixing would just use

ht2_files = Path(snakemake.input.idx).glob("*.ht2*")

instead of

ht2_files = Path(snakemake.input.idx).glob("*.ht2")

... or search for both and combine

However I would rather vote for the user to provide the prefix instead of searching in the provided directory, as problems would arise with multiple indexes in the provided path.

[fgvieira]

I agree. The user providing the prefix sounds safer. Do you think you can make a PR with those changes?

[andreott]

Yes, I will provide a PR

[fgvieira]

Actually, I think it might be better to just provide all the index files as input. Something like:

rule hisat2_align:
    input:
        reads=["reads/{sample}_R1.fastq", "reads/{sample}_R2.fastq"],
        idx=multiext("index/ref", ".1.ht2", ".2.ht2", ".3.ht2", ".4.ht2", ".5.ht2", ".6.ht2", ".7.ht2", ".8.ht2"),
[...]

Then, I think you just need to delete line:

snakemake-wrappers/bio/hisat2/align/wrapper.py

Line 29 in caa3b3f

ht2_files = Path(snakemake.input.idx).glob("*.ht2")

And fix line:

snakemake-wrappers/bio/hisat2/align/wrapper.py

Line 30 in caa3b3f

idx_prefix = os.path.commonprefix(list(ht2_files)).rstrip(".")

[andreott]

I think the problem is, if you build the index within the workflow, you will not know in advance if it is a large index.

[fgvieira]

Unless you force hisat2-build to build a large index.

[andreott]

Hi, getting back to this one - sorry was quite busy the last weeks. Well for me, both are ok. As for the STAR wrapper, I guess it would be more portable not to provide the exact filenames and forcing the user to build the large index to enforce portability. Also I realised that my previous concern (multiple indices in the same directory) is not valid, as the wrapper is supposed to be called with the directory as output, so it will be overwritten by the next indexing run.