Geodata object - Githubissues

mbaudis commented 7 years ago

This issue is to get feedback on how to implement geodata support in the GA4GH schema. Principle options would be:

adoption of the RFC 7946 "GeoJSON" specification (would this have to be replicated in the schema?)
creation of a generic prototype object in metadata.proto, representing a limited subset of GeoJSON attributes (i.e. point object; see this AVRO representation in the deprecated metadata development branch)

Feedback/comments appreciated!

david4096 commented 7 years ago

Hi Michael!

Leaving this here for now https://github.com/mapbox/geobuf

Do you have an example of data to be interchanged that includes geospatial values?

mbaudis commented 7 years ago

David @david4096,

sure there are a number of ways to include/reference/encode geodata; but I rather would reference/track a standard (one could use the Google maps API specs, too, but I'd prefer IETF or ISO to some arbitrary commercial implementer)
overhead is minimal for points in GeoJSON (in contrast to country/regional borders - that's overdoing denormalisation, even for me)
as for examples: every GEO sample (should) include the submitter's city ... at least, which is a pain if done in free text (Paris, Texas ...)
as another example, for arraymap.org | progenetix.org, we attempt to map/refine to get geographic sample provenance, to link analyses & results to possible environmental or genetic background bias
and for non human samples, there are e.g. these environmental sampling projects

So, definitely, many/most first line data objects (biosample, individual, experiment ...) can make use of geodata; and this should be provided in a format maintained outside of GA4GH (like ISO8601 for times).

david4096 commented 7 years ago

@mbaudis Agreed on all accounts, and thank you for the link, fascinating work.

I linked to mapbox's geobuf, which is a compression of geojson using protocol buffers. My hope is that we can leverage their work to rapidly iterate on this feature.

Let's constrain this issue to simply finding an interchange model for these data. If queries become necessary for basic interchange, we can address that feature in those services. For now, we ought to be able to add a GeoJSON-like field to the messages you've mentioned. Here is a link to mapbox's geobuf proto.

david4096 commented 7 years ago

Should this be added as a named field to any messages? The location of the sequencing center for an Experiment message?

mbaudis commented 7 years ago

Could be a named field, if only considering one per record; or carry a label (e.g. listing multiple contacts, collaborators ...).

mbaudis commented 7 years ago

Please follow up on this in PR https://github.com/ga4gh/ga4gh-schemas/pull/864 .

ga4gh / ga4gh-schemas

Geodata object #771