Support build 38

[pjvandehaar]

• [x] allow the user to set the build in config.py.
  • For now, just set build = '38' in config.py, and leave it undocumented so we can change it.
  • Later: I wish that build and fields were more explicit. For association-file-parsing, I'd prefer to auto-generate the parser config and write it to ./generated-by-pheweb/parser-config.json. Build should go with it. Tooltip config, though closely related, should probably go somewhere else.
• [x] Use the appropriate gencode and dbsnp.
  • Files affected: <cache>/gene_aliases.marisa_trie <cache>/genes-27.bed <cache>/rsids-150.vcf.gz and several temp files
  • [ ] We could make file_utils.common_filepaths['genes'] magically handle the version. Sure, sounds interesting, do it.
  • [ ] Download gencode from ftp://ftp.ebi.ac.uk/pub/databases/gencode/Gencode_human/release_28/, and set file_utils.genes_version = '28'. In theory this could be different for different builds, but for now it's the same so don't bother dicting it.
  • [ ] Make sure the gencode parser works for both builds.
  • [ ] Download dbsnp from ftp://ftp.ncbi.nlm.nih.gov/snp/organisms/human_9606_b151_GRCh37p13/VCF/ and ftp://ftp.ncbi.nlm.nih.gov/snp/organisms/human_9606_b151_GRCh38p7/VCF/. Set file_utils.dbsnp_version = '151'. Makepheweb.load.download_rsids` deal with the URL difference.
  • [ ] Make sure the dbsnp parser works for both builds.
• [x] Javascript: Set window.model.genome_build. Adjust LZ's datasources for LD, recombination rates, and genes.
  • [ ] Ask the appropriate people where to get these data.
  • [ ] For genes, perhaps each pheweb should just use its own local API.
  • [ ] Once this works, remove window.model.genome_build and instead directly pass it to each function that needs it.
• [x] Jinja: show the build on the about page.

[abought]

For LZ.js datasources (the portaldev web API), see discussion on the #Locuszoom channel in our slack today. Basically, the current status is as follows:

• Genes already have build 38 data
• There is a path to LD data, which we should coordinate with Daniel (LD server) and Ryan (portal API integration) when Daniel gets back
• Surprisingly, recombination rates lack a modern dataset; liftover may be an option

If you haven't looked at the required LZ updates yet, feel free to create a ticket in that repo and assign me as help needed. We can discuss this (and GWAS catalog) tomorrow.

[abought]

Adding relevant PR link: https://github.com/statgen/locuszoom/pull/149

This PR will add LD and gene support for h38. It's a work in progress, as we're still beta testing the new LD server a bit (see checklist for plans and future directions).

[abought]

For the Locuszoom side of this task:

Build 38 support has been merged into LZ.js develop branch, including use of new LD and Recomb rate API data. (thanks, Ryan and Daniel!) Gene data was already available.

If you specify genome_build: 'GRCh38' in plot.state, all LZ.js data sources will automatically choose a dataset appropriate to that build. Would this help reduce the amount of custom code needed for your configuration?

Here is an example showing what that looks like (notice that most sources no longer encode a source: value of their own) https://github.com/statgen/locuszoom/blob/develop/examples/gwas_catalog.html#L86-L111

[abought]

To help track work and blockers: the separate genelocator package handles build38 .

I've also got some RSID parsers that operate on the new DB snp build 153 format, for both genome builds 37 and 38. Let me know what would be useful to get this moving.