Allow Varcode to work with mouse data via Genome

[tavinathanson]

This is a quick fix to allow Varcode to work with mouse data (and an initial test for that) by accepting Genome objects.

It's certainly confusing to be using ensembl and genome somewhat interchangeably here. I'd like a follow-up PR (and will create an issue, if this order-of-things sounds good to you) to liberate Varcode from Ensembl much like openvax/pyensembl#99, but for now this seems useful.

[timodonnell]

Many thanks for doing this in a way that does not break all of our existing code (which passes in ensembl_version to load_vcf)!

[tavinathanson]

@timodonnell Fewer tests for me to fix, too! The intent of openvax/pyensembl#99 is also to keep the EnsemblRelease API intact despite refactoring the innards and adding a superclass to it.

[iskandr]

Finding lots of little bugs in the PyEnsembl PR we merged, but also finding a design flaw wherein each GenomeSource only tells you the command you need to download e.g. a protein FASTA but that command can't be run in isolation (since the pyensembl script always requires a GTF).

I guess we'll figure out a better design when you come back.

[iskandr]

To clarify on the last comment, to install the mouse genome you need to run:

pyensembl install --reference-name "GRCm38" --transcript-fasta-path-or-url "ftp://ftp.ensembl.org/pub/release-81/fasta/mus_musculus/cdna/Mus_musculus.GRCm38.cdna.all.fa.gz" --protein-fasta-path-or-url "ftp://ftp.ensembl.org/pub/release-81/fasta/mus_musculus/pep/Mus_musculus.GRCm38.pep.all.fa.gz" --gtf-path-or-url "ftp://ftp.ensembl.org/pub/release-81/gtf/mus_musculus/Mus_musculus.GRCm38.81.gtf.gz"

but if this genome is missing you'll see errors like:

nose.proxy.ValueError: Genome sequence data (GenomeSource(transcript_fasta_path_or_url=ftp://ftp.ensembl.org/pub/release-81/fasta/mus_musculus/cdna/Mus_musculus.GRCm38.cdna.all.fa.gz)) is not currently installed for this genome source. Run pyensembl install --transcript_fasta_path_or_url "ftp://ftp.ensembl.org/pub/release-81/fasta/mus_musculus/cdna/Mus_musculus.GRCm38.cdna.all.fa.gz" or call Genome(reference_name="GRCm38", transcript_fasta_path_or_url="ftp://ftp.ensembl.org/pub/release-81/fasta/mus_musculus/cdna/Mus_musculus.GRCm38.cdna.all.fa.gz")).install()

[iskandr]

Additionally, it looks like EnsemblReleaseSource no longer gets used and thus even human Ensembl references emit an error like the one above.

[tavinathanson]

Offline discussion re the design flaw: it still "works", but might be an annoyance to run multiple commands. We agreed to table that until openvax/pyensembl#100 is addressed.

[tavinathanson]

Re the bugs: @iskandr and I worked through them together and they're summarized/addressed here: openvax/pyensembl#108

Offline discussion re EnsemblReleaseSource: that was intended to be used, and is used in the PR above.

[tavinathanson]

@iskandr I don't remember how we left it with this PR, since it's hacky but not entirely awful. What would you like to see before we merge?

[iskandr]

@tavinathanson Let's merge this, and get mouse epitope predictions out of Topiary, and figure out a better organization next week.