Add ability to output ClinVar XML

[apriltuesday]

Most of this diff is just splitting the single clinvar_xml_io.py into multiple files, one for each class. The actual functionality change is in clinvar_dataset.py and its test.

I also had to update the consequences for SNP and structural variants in the end-to-end test to match updates to VEP, please double-check those as well (can make a separate PR if you prefer).

[coveralls]

Pull Request Test Coverage Report for Build 4245543983

• 297 of 324 (91.67%) changed or added relevant lines in 7 files are covered.
• 1 unchanged line in 1 file lost coverage.
• Overall coverage increased (+1.02%) to 83.869%

---

Changes Missing Coverage	Covered Lines	Changed/Added Lines	%
eva_cttv_pipeline/clinvar_xml_io/clinvar_xml_io/xml_parsing.py	22	23	95.65%
eva_cttv_pipeline/clinvar_xml_io/clinvar_xml_io/clinvar_record.py	64	68	94.12%
eva_cttv_pipeline/clinvar_xml_io/clinvar_xml_io/clinvar_trait.py	45	50	90.0%
eva_cttv_pipeline/clinvar_xml_io/clinvar_xml_io/clinvar_measure.py	124	141	87.94%

Files with Coverage Reduction	New Missed Lines	%
eva_cttv_pipeline/trait_mapping/ols.py	1	87.8%

Totals
Change from base Build 4026072309:	1.02%
Covered Lines:	1305
Relevant Lines:	1556

---

💛 - Coveralls

[tcezard]

That looks good.
I understand that this allows to write the XML not to make addition/changes to it yet.
The consequence changes are strange and I will report them to Ensembl. They affect large deletions that overlap with large portion of genes so splice_polypyrimidine_tract_variant was wrong but coding_sequence_variant is not any better. I guess they can't be saying transcript_ablation since it's only a partial ablations.

[tcezard]

I was curious so I looked into one of the consequence in more detail:
nsv1197494 overlaps the second half of IRAK1 deleting several exon/intro.
We report it as a coding_sequence_variant but VEP reports it as

        "consequence_terms": [
          "coding_sequence_variant",
          "5_prime_UTR_variant",
          "intron_variant",
          "feature_truncation"
        ]

Maybe we need to revise how we're reporting consequences

[apriltuesday]

@tcezard I've created #366 for this so we can come back to it, thanks for raising. Feel free to edit the issue with any additional details.

[M-casado]

Overall I took a look at the slim clinvar_dataset.py, for which I left a quick question (since I don't know the insides of the classes).

Re. the consequence changes, similar to @tcezard, I took a look at one of the examples: ENSG00000044524 (EPHA3) (89335416_89368021del) that changes, like most, from splice_polypyrimidine_tract_variant to coding_sequence_variant. In this case I think, without taking into account the insides of VEP, the change makes sense: the deletion spans 32Kb and engulfs 2 exons, so I assume that the most severe consequence is coding_sequence_variant out of the ones VEP reports (and above splice_polypyrimidine_tract_variant)
Summary of VEP's:

{
  "transcript_consequences": [
  "coding_sequence_variant",
  "intron_variant",
  "feature_truncation"
   ]
}

[M-casado]

I assume test_output.xml.gz is the annotated version of ClinVar's that we would also distribute some how in case people want to use it, right?

[M-casado]

Or are we missing the annotation chunk in between?

[apriltuesday]

Yes that's right, there's no annotation yet and we're just outputting the raw RCV XML for each record. The annotation will be coming in a subsequent PR.