🍱 Re-curating cellxgene: Annotating with features values & mapping version families

[falexwolf]

I wanted to put the following on a slide in the demo deck, but I realize that it's falling short because it doesn't associate labels with features.

https://lamin.ai/laminlabs/cellxgene/artifact/Ntdynj9uX3LGLRCK2xcV

Can we tweak the curation script so that features are mapped? And also annotate with feature values?

Also: Can we map version families to collapse the following search result? See this Slack thread for context.

[Zethson]

Can we split this into two issues?

[falexwolf]

Yes, we can split this into two issues!

I thought I'd put into one as these should both be 3 lines of code in the curation script/notebook, so maybe a single rather quick PR.

[falexwolf]

To clarify what I mean, I'll now use the example from our scRNA guide for the slide.

https://docs.lamin.ai/scrna

https://lamin.ai/laminlabs/lamindata/artifact/xJkeL0OxEFIpvGWKdpne

And we should double check that everything makes sense for this one before doing a bulk-re-curation.

On the hub, this one also looks somewhat strange. Not clear if this is something potentially going wrong in the frontend or whether there is still weirdness in feature tracking.

[Zethson]

if field in ["donor_id", "suspension_type", "tissue_type"]:
            records = orm.from_values(terms, field="name")
            if len(records) > 0:
                # stratify by feature so that link tables records are written
                artifact.labels.add(records, feature=getattr(features, field))

In the census registration script must be adapted by removing the if statement.