2397 ns forest markers

[dosumis]

Turning this into a draft PR as a place for comments and discussion

@aleixpuigb - hope that's OK.

[dosumis]

ERROR:root:Failed to load pattern file: ../patterns/dosdp-patterns/MarkersToCells.yaml

[aleixpuigb]

The Marker to cell pattern is not finished, I will leave it in draft mode while I solve the issues.

[dosumis]

@hkir-dev - Could you co-ordinate with @aleixpuigb on this. High priority. We will also need these patterns in the CL-KG build. We need something very similar to BDSO.

Main issue:

• How to efficiently co-ordinate axioms needed for term partly defined by NS-Forest markers & the marker set term. I guess we used some Python in BDSO. Might be best to have this here so we're not relying on duplicating content between TSVs.

Specific issues:

1. We need a slot for xrefs. These should take the DOIs for notebooks detailing the analysis. Maybe we could add this as an xref on some text in the comment field on the marker set.
2. Type marker set as "sequence_collection": "SO:0001260"
3. Marker set needs regular var for cell type (range "CL:0000000")
4. We have a new AP for fbetaConfidenceScore: "STATO:0001000"
5. Patterns have the same pattern_name

[hkir-dev]

Now the marker urls are generated like http://purl.obolibrary.org/obo/entrez_286133 but in the BDSO it is in the form of http://identifiers.org/ncbigene/286133 which is resolvable.

To do this we need to override the make file rules that run $(DOSDPT) generate command and provide a prefix mapping file to it: --prefixes=template_prefixes.yaml

[dosumis]

We will also need an import file for these to get labels. (I think we had ensembl & NCBI for BDSO). I'm wondering whether, longer term, a more sustainable approach would be to have a separate repo for generating marker objects and make imports from that to the KG and CL.

[hkir-dev]

Recently STATO added a new term for F Beta Score ISA-tools/stato#86 (comment).

Should we use STATO:0000663(F-beta score) instead of fbetaConfidenceScore: "PCL:0010062"

[dosumis]

yes please

[hkir-dev]

I don't think this pattern is running at all. I couldn't see any terms generated by this template in the definitions.owl. I couldn't see any terms in my local build as well.
Errors are:

• missing column for pattern variable <Marker_set>
• missing column for pattern variable <Marker_set_xref>
• missing column for pattern variable <has_characterization_set>

[hkir-dev]

How to efficiently co-ordinate axioms needed for term partly defined by NS-Forest markers & the marker set term. I guess we used some Python in BDSO. Might be best to have this here so we're not relying on duplicating content between TSVs.

Yes, in BDSO, we have a manual step before the ODK prepare_release that reads the source files and programmatically generates the template data files. I didn't notice any script modifications in the CL repo, which makes me hesitant to add these scripts as it could complicate the maintenance of the CL build pipeline. However, if the number of markers will significantly increase in the future and making programmatic management inevitable, I can propose two solutions:

1. Create a new source file at src/patterns/data/raw/NSForestMarkersSource.tsv programmatically populate the contents of MarkersToCells.tsv and NSForestMarkers.tsv before executing the dosdp templates.
2. Modify the makefile to include a goal similar to the existing goal to download a dosdp template from a google sheet (though this goal seems inactive currently). Similarly, we can manage markers in a separate repo and pull them at runtime.

[dillerm]

Specific issues:

1. We need a slot for xrefs.  These should take the DOIs for notebooks detailing the analysis.  Maybe we could add this as an xref on some text in the comment field on the marker set.

Is this referring to GO's hasDbXref annotation? I'm wondering if IAO's 'definition source' (IAO:0000119) might be more appropriate here.

[dosumis]

This has been superseded by work on https://github.com/Cellular-Semantics/CellMark/tree/main - which will be a source of imports for for CL as well as a source a source of markers more broadly for https://github.com/Cellular-Semantics/CL_KG

[dosumis]

Is this referring to GO's hasDbXref annotation? I'm wondering if IAO's 'definition source' (IAO:0000119) might be more appropriate here.

Using GO's hasDbXref aligns better with current CL standards.