Incorporating Whole Mouse Brain subclass terms into CL
#2809
Comments
|
Can you create a script to convert the cell_label(s) into more readable names? Then we can keep the cell_label entries as exact synonyms. |
|
Renee's team have completed a pretty comprehensive analysis of marker combinations for mouse brain subclasses that are derived from NS-Forest, the Yao paper, and the Zhang paper - |
Yep. She has shared & we're working on folding them in |
Already have a ticket for that: |
|
We will v.likely end up adding a more complete set of Mouse Brain T-type terms to CL in early 2025. We will host a version with PCL IDs on an OLS as preliminary for review, migrating to CL IDs once we have some agreement in place. https://github.com/orgs/Cellular-Semantics/projects/4 |
STATUS: DRAFT FOR DISCUSSION
Summary
BICAN/HMBA are pushing very hard to start incorporating cell types defined by transcriptomics - starting with
subclasslevel terms from the whole mouse brain - at least for neurons, possibly for some other neural cell types. See details in fold at end of this post. This will set precedent for incorporating a deeper transcriptomic hierarchy at a later date. Possible timing - early 2025 - as there may be some adjustments by then. For now we are folding into PCL.Challenges:
Folding in transcriptomic hierarchy as SubClassOf is potentially dangerous because it may not fit with property inheritance (initial phase will not include this hierarchy but later phases will). We could use a new relation for transcriptomic hierarchy - strictly this should be an AP for class-class relationships but this wouldn't show up in standard ontology browsers. Would hacking with an OP be acceptable?
Initial terms will be for mouse, but we will soon have putatively homologous cell types for human, marmoset and macaque + cross species mappings.
What we have to work with:
Metadata
We have some level of mapping to CL. We should keep original names as symbols but these can potentially be expanded into more readable labels - with brain region, Neurotransmitter, and sometimes marker and occasionally other semantic load (e.g. cortical layer, projection pattern) coming from annotation transfer. It is important to note that the brain region mentioned in the name is a major brain region for that cell type - it may not apply to all subclasses and does not encompass all the regions where it is located. The same may also be true of markers in names - although this is testable. Other semantic load within names may also not cleanly inherit - in many cases we just don't know.
There is specification of markers for identifying each cell type (combo.markers in table below) - which may be consistent down the hierarchy (testable) - and a more complete accounting of location using MBA based on spatial transcriptomics (MerFish) at the cluster level - a couple of levels down from subClassOf. (It may be possible to summarise this up the MBA hierarchy (?))
Pipelines
We will have pipelines in place for autogenerating these terms from metadata - built on the BDSO pipeline (ODK + LinkML + templates + some Python glue). Taking shape on https://github.com/Cellular-Semantics/whole_mouse_brain_ontology/tree/version2
(The LinkML component comes from CAS https://github.com/Cellular-Semantics/cell-annotation-schema)
WMB SubClass table
annotation.csv
The text was updated successfully, but these errors were encountered: