[Design pattern] Develop design pattern for amacrine cell terms and definitions

[scheuerm]

@aleixpuigb and I are working with the eye cell atlas group to annotate the cell types for version 1 of the atlas. The final single cell data is in hand (https://cellxgene.cziscience.com/collections/4c6eaf5c-6d57-4c76-b1e9-60df8c655f1e ) and the manuscript is being prepared. While many of the cell types will have labels that are easily recognizable by the community and will link to prior knowledge/known biology (OFF midget ganglion cell, diffuse bipolar 1 cell), the amacrine cell subclasses, of which there are 95, are currently labelled as HAC1 - HAC95. I'd like to consider:

1. alternative design patterns for these terms, together with the advantage and disadvantages from a semantic perspective, to present to the community for them to decide which they prefer.
2. A design pattern for how these cell types will be defined in the ontology.

Some initial thoughts:

• Useful components of a new cell type term name:
• Retain connection with prior publications, either in the name or through the use of synonyms
• Include a useful characteristics beyond a simple sequential number, e.g., a characteristic marker gene, for ease of use. The NS-Forest markers for the amacrine cells are pretty good, with median Fbeta0.5 = 0.65 and median precision = 0.83 for the optimal combinations.

HAC66 cell (CL:XXX) =(def) An amacrine cell that selectively expresses the USH2A transcript in human.
Term option 1 - HAC66
Term option 2 - HAC66 USH2A
Term option 3 - USH2A HAC66
Term option 4 - USH2A-expressing AC
Term option 5 - USH2A-expressing HAC66

If there is a single defining marker, then all of these options would be reasonable.

HAC2 cell (CL:YYY) =(def) An amacrine cell that selectively expresses the combination of TSHZ2 and TMEM132D transcripts in human.
Term option 1 - HAC2
Term option 2 - HAC2 TSHZ2
Term option 3 - TSHZ2 HAC2 
Term option 4 - TSHZ2-expressing AC
Term option 5 - TSHZ2-expressing HAC2
Term option 6 - HAC2 TSHZ2 and TMEM132D

If there are multiple defining markers, using a single marker in the name may not reflect the actual specificity of expression since it may also be expressed in other cell types. If all of the required markers are included in the name, the name can become fairly long. For most of the HAC subtypes 2 - 3 markers are selected, but there are a couple where 5 or 6 markers were selected.

[dosumis]

Some thoughts:

1. We will need full names (e.g. human amacrine cell 66) but can have official symbols too (e.g. HAC66)
2. While we can and do accept numbers in names, we need to be very careful of potential clashes. Another group numbering human amacrine cells 1 - n would end up giving unrelated cells the same name.
3. If we decide not to use numbers, I favour choosing one gene name to extend the name - ensuring that the results are unique. Anything else will => something too long. It is easy to write a script that achieves this. The full list of genes should go in the definition.
4. The model organism databases all forbid gene symbols abbreviations starting with the first letter of the species name, e.g. FlyBase has no gene symbols starting with D where the D stands for Drosophila. We may want to consider a similar rule for cell type symbols (we will always capture external usage in symbols).