Decide on groupings for all diagnoses

[allyhawkins]

For figure 1, we will want to create different groups for all the diagnoses so that we can facet and color by group. This issue should be used for a discussion to figure out the following two things:

1. How to divide the diagnoses. I will post a table with diagnoses in one column and group in the other column for everyone to check and decide on how exactly we want to label each group. One thought is we should use the ontology ID's for diagnosis instead of the submitter provided diagnoses?
2. We will need to decide on a color palette and which color to use for each diagnoses.

[allyhawkins]

Below I have taken the table we use to assign ontology ID's and assign each diagnosis to a group or category that we can use for plotting. I was a bit more specific about the names used for the groups so used Leukemia instead of Blood and Brain and CNS instead of Brain.

1. Do we want to display the human_readable_value associated with the ontology ID's in the figures or the submitted_diagnosis. Originally I thought the ontology human readable value, but then we don't display that on the front end of the portal so if someone went to go look for a specific sample, I think that would get confusing.
2. What do we do about the Normal/non-cancerous samples? I don't think they really belong in a category, so I wonder if we want to not include them in the grouped plot, but maybe as Josh suggested we show the bar plot of every single diagnosis in the supplementals and in that plot we could include the normal.
3. If we only have four groups here then are we okay with picking four colors from the Okabe-Ito color palette? Does anyone have any favorite color palettes that we could choose from? I do not think we should be coloring every single diagnosis.

Tagging @jaclyn-taroni @jashapiro @sjspielman for any thoughts.

ontology_id	human_readable_value	submitted_diagnosis	group
MONDO:0016684	Anaplastic astrocytoma	Anaplastic astrocytoma	Brain and CNS
MONDO:0016700	Anaplastic ependymoma	Anaplastic ependymoma	Brain and CNS
MONDO:0016734	Anaplastic ganglioglioma	Anaplastic ganglioglioma	Brain and CNS
MONDO:0021640	Grade III glioma	Anaplastic glioma	Brain and CNS
MONDO:0020560	Atypical teratoid rhabdoid tumor	Atypical teratoid rhabdoid tumor	Brain and CNS
MONDO:0022965	Desmoplastic infantile ganglioglioma	Desmoplastic ganglioglioma	Brain and CNS
MONDO:0005499	Brain glioma	Diffuse midline glioma	Brain and CNS
MONDO:0005505	Dysembryoplastic neuroepithelial tumor	Dysembryoplastic neuroepithelial tumor	Brain and CNS
MONDO:0019002	Lhermitte-Duclos disease	Dysplastic gangliocytoma	Brain and CNS
MONDO:0016698	Ependymoma	Ependymoma	Brain and CNS
MONDO:0019009	Isolated focal cortical dysplasia	Focal cortical dysplasia	Brain and CNS
MONDO:0016733	Ganglioglioma	Ganglioglioma	Brain and CNS
MONDO:0021193	Neuroepithelial neoplasm	Ganglioglioma/ATRT	Brain and CNS
MONDO:0016729	Mixed neuronal-glial tumor	Glial-neuronal tumor	Brain and CNS
MONDO:0018177	Glioblastoma	Glioblastoma	Brain and CNS
MONDO:0100342	Malignant glioma	High-grade glioma	Brain and CNS
MONDO:0858940	Infant-type hemispheric glioma	Infant-type hemispheric glioma	Brain and CNS
MONDO:0021637	Low grade glioma	Low-grade glioma	Brain and CNS
MONDO:0007959	Medulloblastoma	Medulloblastoma	Brain and CNS
MONDO:0016699	Myxopapillary ependymoma	Myxopapillary ependymoma	Brain and CNS
MONDO:0016691	Pilocytic astrocyoma	Pilocytic astrocytoma	Brain and CNS
MONDO:0016692	Pilomyxoid astrocytoma	Pilomyxoid astrocytoma	Brain and CNS
MONDO:0016690	Pleomorphic xanthoastrocytoma	Pleomorphic xanthoastrocytoma	Brain and CNS
MONDO:0000640	Central nervous system primitive neuroectodermal neoplasm	Primitive neuroectodermal tumor	Brain and CNS
MONDO:0002546	Schwannoma	Schwannoma	Brain and CNS
MONDO:0007667	Subependymoma	Subependymoma	Brain and CNS
MONDO:0018874	Acute myeloid leukemia	Acute myeloid leukemia	Leukemia
MONDO:0004947	B-cell acute lymphoblastic leukemia	B-cell acute lymphoblastic leukemia	Leukemia
MONDO:0100291	Early T-cell progenitor acute lymphoblastic leukemia	Early T-cell precursor T-cell acute lymphoblastic leukemia	Leukemia
MONDO:0020743	Mixed phenotype acute leukemia	Mixed phenotype acute leukemia	Leukemia
MONDO:0004963	T-cell acute lymphoblastic leukemia	Non-early T-cell precursor T-cell acute lymphoblastic leukemia	Leukemia
MONDO:0004963	T-cell acute lymphoblastic leukemia	T-cell acute lymphoblastic leukemia	Leukemia
MONDO:0020743	Mixed phenotype acute leukemia	T-myeloid mixed phenotype acute leukemia	Leukemia
MONDO:0006639	Adrenal cortex carcinoma	Adrenocortical carcinoma	Other solid tumors
MONDO:0011719	Gastrointestinal stromal tumor	Gastrointestinal stromal tumor	Other solid tumors
MONDO:0005040	Germ cell tumor	Germ cell tumor	Other solid tumors
MONDO:0008380	Retinoblastoma	Retinoblastoma	Other solid tumors
MONDO:0002728	Rhabdoid tumor	Rhabdoid tumor	Other solid tumors
MONDO:0006058	Wilms tumor	Wilms tumor	Other solid tumors
MONDO:0005072	Neuroblastoma	Neuroblastoma	Other solid tumors
MONDO:0002926	Clear cell sarcoma	Clear cell sarcoma	Sarcoma
MONDO:0005006	Clear cell sarcoma of the kidney	Clear cell sarcoma of the kidney	Sarcoma
MONDO:0019373	Desmoplastic small round cell tumor	Desmoplastic small round cell tumor	Sarcoma
MONDO:0015795	Undifferentiated embryonal sarcoma of the liver	Embryonal sarcoma	Sarcoma
MONDO:0017387	Epithelioid sarcoma	Epithelioid sarcoma	Sarcoma
MONDO:0012817	Ewing sarcoma	Ewing sarcoma	Sarcoma
MONDO:0004557	Congenital fibrosarcoma	Infantile fibrosarcoma	Sarcoma
MONDO:0009807	Osteosarcoma	Osteosarcoma	Sarcoma
MONDO:0005212	Rhabdomyosarcoma	Rhabdomyosarcoma	Sarcoma
MONDO:0002927	Spindle cell sarcoma	Spindle sarcoma	Sarcoma
MONDO:0010434	Synovial sarcoma	Synovial sarcoma	Sarcoma
MONDO:0020661	Undifferentiated round cell sarcoma	Undifferentiated round cell sarcoma	Sarcoma
PATO:0000461	Normal	Non-cancerous

[sjspielman]

maybe as Josh suggested we show the bar plot of every single diagnosis in the supplementals and in that plot we could include the normal.

This makes sense to me.

Originally I thought the ontology human readable value, but then we don't display that on the front end of the portal so if someone went to go look for a specific sample, I think that would get confusing.

How about we match the portal in the main text, but toss a ontology human-readable version of the same figure into the supp? Edit - maybe this is the same figure as the normal/non-cancerous go into?

If we only have four groups here then are we okay with picking four colors from the Okabe-Ito color palette? Does anyone have any favorite color palettes that we could choose from?

I don't know about using Okabe-Ito, specifically/only because we might want to use Okabe-Ito for the UMAPs. If we don't want to reserve this palette for UMAPs then it definitely seems fine to use here! In terms of other palettes, I generally like Dark2 or a discrete version of the regular viridis, but I really don't feel strongly either way.

[allyhawkins]

I don't know about using Okabe-Ito, specifically/only because we might want to use Okabe-Ito for the UMAPs. If we don't want to reserve this palette for UMAPs then it definitely seems fine to use here! In terms of other palettes, I generally like Dark2 or a discrete version of the regular viridis, but I really don't feel strongly either way.

I think this is a really good point! If we want to avoid overlapping palettes with any future figures, then we probably want to pick from a palette we aren't likely to use again. I feel like Dark2 could fall into that category too... What about the first 4 colors I circled from Set1?

Another thought, we could pick 4 colors from the cancer group colors that were used in OpenPBTA...

[jaclyn-taroni]

To comment on color palette only:

I don't believe Dark2 or Set1 are particularly accessible, and the yellow in viridis ends up low contrast with a white background, in my opinion:

What about hcl.colors(4, "cividis"): #00214E #545C71 #A89F76 #FFE93F

If we're worried about the contrast with the brightest yellow, we could do hcl.colors(5, "cividis")[1:4]: #00214E #3E4C6E #7C7C7C #BFB170

Or hcl.colors(5, "batlow")[1:4]: #201158 #005E5E #578B21 #E89E6B

Or palette.colors(n = 5, palette = "R4")[-1]: #DF536B #61D04F #2297E6 #28E2E5

Regardless of the palette we choose, we should try to encode this information with facets or panels as well!

Edited to add: I would include a grey color for normals in this palette. The worst thing that happens is you don't use it.

[sjspielman]

I don't believe Dark2 or Set1 are particularly accessible,

Just noting that of the brewer palettes, only Dark2, Set2, and Paired are CVD-friendly. We should try to stick with something that is generally accessible, but I agree that Dark2 can sometimes be hard to see!

[jaclyn-taroni]

I'm going to respond to @allyhawkins 3 points all in one place.

1. I would use submitted_diagnosis for the reason you give – we want people to be able to reference data from the Portal.

2. Add a non-cancerous group. Pick a grey color for non-cancerous samples. We'll want to check it for contrast with whatever the other 4 groups' colors end up being. If we don't end up needing it, there's very little harm in having it!

3. I compiled my thoughts on palettes here https://github.com/AlexsLemonade/scpca-paper-figures/issues/10#issuecomment-1863456338. And to record something discussed on Slack:

   Just noting that of the brewer palettes, only Dark2, Set2, and Paired are CVD-friendly. We should try to stick with something that is generally accessible, but I agree that Dark2 can sometimes be hard to see!

   Once you are at n = 4, only Paired meets this criterion.

   We should prioritize a palette with darker colors because, if we plan to use it throughout the paper, we'll want the flexibility to color points on a white background.

[allyhawkins]

Closing this and any additional discussion can happen in #20