Add modality

[BAevermann]

Design (@brianraymor)

obs

...

modality

Key	modality
Annotator	Curator MUST annotate.
Value	categorical with str categories. This MUST be "epigenomics" or "transcriptomics". This MUST be the correct type for the corresponding assay: For Assay MUST Use 10x multiome [EFO:0030059] "epigenomics" or "transcriptomics" 10x scATAC-seq [EFO:0030007] "epigenomics" 10x transcription profiling [EFO:0030080] and its descendants "transcriptomics" BD Rhapsody Targeted mRNA [EFO:0700004] "transcriptomics" BD Rhapsody Whole Transcriptome Analysis [EFO:0700003] "transcriptomics" CEL-seq2 [EFO:0010010] and its descendants "transcriptomics" DroNc-seq [EFO:0008720] "transcriptomics" Drop-seq [EFO:0008722] "transcriptomics" GEXSCOPE technology [EFO:0700011] "transcriptomics" inDrop [EFO:0008780] "transcriptomics" MARS-seq [EFO:0008796] "transcriptomics" mCT-seq [EFO:0030060] "epigenomics" or"transcriptomics" MERFISH [EFO:0008992] "transcriptomics" methylation profiling by high throughput sequencing [EFO:0002761] and its descendants "epigenomics" microwell-seq [EFO:0030002] "transcriptomics" Patch-seq [EFO:0008853] "transcriptomics" ScaleBio single cell RNA sequencing [EFO:0022490] "transcriptomics" scATAC-seq [EFO:0010891] "epigenomics" sci-Plex[EFO:0030026] "transcriptomics" sci-RNA-seq [EFO:0010550] and its descendants "transcriptomics" Seq-Well [EFO:0008919] and its descendants "transcriptomics" Smart-like [EFO:0010184] and its descendants "transcriptomics" spatial transcriptomics [EFO:0008994] and its descendants "transcriptomics" SPLiT-seq [EFO:0009919] "transcriptomics" STRT-seq [EFO:0008953] "transcriptomics" TruDrop [EFO:0700010] "transcriptomics" If the assay does not appear in this table, the most appropriate value MUST be selected and the curation team informed during submission so that the assay can be added to the table.

Context

At current the CELLxGENE schema does not capture the concept of a detected analyte.

This concept is usually implied in the assay name, for example the mRNA analyte as detected by 10x 3' transcriptional profiling. However, for assays such as "10x multiome" the analyte detected is ambiguous as it measures both mRNA and open chromatin.

This distinction is required by downstream tools such as Census or Expression to filter supported vs unsupported data.

[brianraymor]

~Would this only be required when there was complete support for 10X multiome? Or is this being proposed to allow census to consume this assay now and be "future proofed" when/if ATAC is supported?~

~Reviewing the draft for the assay tier proposal:~

~1. 10X multiome (RNA) is experimental. And there are 60 datasets?~ ~2. 10X multiome (ATAC) is unsupported.~

~How is 10X multiome currently modeled?~

[jahilton]

~Will also flag mCT-seq which measures both RNA & methylation (we have 1 Collection, which holds the expression data)~

~In Lattice, we use biological_macromolecule with an enum of RNA,DNA, or protein~ ~As an alternative, I could also imagine a field that lists the type of measurement being represented (expression, accessibility, etc.)~

[BAevermann]

~Assortment of proposals were discussed on 11/30~

~Current proposal~ ~Field name: Modality~ ~Values: (Controlled vocab)~ ~Transcriptomics~ ~Epigenomics~ ~Proteomics~ ~Spatial Transcriptomics~ ~Spatial Proteomics~ ~in-situ hybridization assay~

~Will meet in early Q1 to discuss further.~

[jahilton]

notes the "Modality" proposal:

• to keep the list as short as possible, what's the value in having the 'spatial*' terms? i would anticipate the number of assays in each would be low enough to filter/query on the assay values (or uns.is_single?). spatial could be viewed as an added dimension to any modality, rather than a distinct modality.
• similar, what's the value in 'in-situ hybridization assay'? kind of a different term than the others (more along the lines of plate- vs droplet-based)
• how would we envision capturing AB-derived data (CITE-seq) - is that "proteomics" or a different value?

Based on the proposal, I took a stab at mapping each current assay in the corpus to the Modality values - this sheet - for others to review Biggest Q is that I'm not sure how to characterize the morphology & electrophysiology measurements that are a part of Patch-seq (in addition to the transcriptomics).

[pablo-gar]

Agreed with Jason, we would be overloading this filed with the addition of "spatial". This axis of variation is likely to be already captured by assay. The main goal as I read it is to distinguished between molecules for downstream applications, and with the upcoming support for Spatial, I don't see a need to overlap this variable.

I'd prefer to stick to the name of the molecule or the omics term.

[BAevermann]

Thanks for the mapping @jahilton!

• The "spatial" flavors of were added for future support, so most wouldnt be needed as of right not. In some cases I can see the spatial component as a dimension (ST) as the sequencing readout is the primary result. For the in situ hybridization and antibody based methods, the detection and readout comes during the imaging process and is thereby spatially dependent, i.e., the local density of the cellular compartment will impact the signal detection. You are correct that we can easily use the assay name for the time being as they are limited.
• See reasoning above for why in situ hybrid is its own flavor.
• The CITE-SEQ is an interesting case, as the readout is still sequencing based, mediated by antibody binding. Could splitting these experiments into two assays, one Transcriptomics/ one Proteomics, make it easier to represent in the schema? They are processed together, albeit with special analysis tools ...
• For the patch-seq: we are only acquiring the Transcriptomic component of that assay, right? And we have no plans to acquire the other aspects as they dont currently fit into the concept of CELLxGENE. Can we just ignore?

[jahilton]

• Now we're combining modality with readout. So we would specify "transcriptomics by sequencing" instead of just "transcriptomics"...or we should not overload the modality field & instead capture the readout elsewhere, if needed.
• The CITE-SEQ question is focused on the protein/AB component. 'transcriptomics' is a given. So it sounds like the AB component would be 'proteomics' (we can't say how to capture those because we don't yet know how we'll capture the AB data, in general). And in the spirit of not overloading, a distinction on how that is measured, targeted or unbiased, etc. would be captured elsewhere if needed.
• 👍 to Patch-seq just as 'transcriptomics'

[brianraymor]

April 15 2024 (@BAevermann, @brianraymor, @jahilton, @jychien, @pablo-gar)

obs['modality'] transcriptomics epigenomics proteomics

[brianraymor]

@BAevermann, @jahilton, @jychien, @pablo-gar

Would you please review the draft in the top-level summary comment under Design. (I cannot submit a PR for this field because its schema version is unknown at this time.)

Comments, LGTM, or emojis all accepted. Also feel free to edit in place.

[pablo-gar]

@brianraymor I think it should be "proteomics" for spatial proteomics [EFO:0700000] and its descendants. @jahilton can confirm

otherwise LGTM

[brianraymor]

I think it should be "proteomics

Doh. Cut-n-paste error. Corrected.

[jahilton]

One too many 0 - mCT-seq [EFO:~0~0030060] Need to add "...and its descendants" to sci-RNA-seq [EFO:0010550]. Otherwise, sci-RNA-seq3 is not covered. Could add "...and its descendants" to scATAC-seq [EFO:0010891] and ditch the "10x scATAC-seq [EFO:0030007]" row

Potential risks with relying on descendants for this one. Some hypotheticals:

• If EFO does the appropriate thing and moves 10x multiome under both 10x transcription profiling? Fairly confident that won't happen unless we ask for it.
• If a term is created that is a descendant of both spatial proteomics and spatial transcriptomics

I think we're focused on transcriptomics enough to be resistant to those rare occurrences, but just wanted to raise them.

[brianraymor]

One too many 0 - mCT-seq [EFO:~0~0030060]

Good catch. I owe you a $1. Corrected.

Need to add "...and its descendants" to sci-RNA-seq [EFO:0010550]. Otherwise, sci-RNA-seq3 is not covered.

Added.

Could add "...and its descendants" to scATAC-seq [EFO:0010891] and ditch the "10x scATAC-seq [EFO:0030007]" row

The problem is that 10x multiome is a descendant of scATAC-seq.

Potential risks with relying on descendants for this one. Some hypotheticals:

• If EFO does the appropriate thing and moves 10x multiome under both 10x transcription profiling? Fairly confident that won't happen unless we ask for it.
• If a term is created that is a descendant of both spatial proteomics and spatial transcriptomics I think we're focused on transcriptomics enough to be resistant to those rare occurrences, but just wanted to raise them.

This could be part of the review when the schema updates EFO in a version?

[jychien]

All the spatial assays are represented as descendants of either spatial transcriptomics or spatial proteomics. Is finer granularity is required?

Not that we're accepting this assay, but FYI, NanoString digital spatial profiling is a child of spatial transcriptomics and it has in its definition 'spatial analysis of RNA and protein'. Other than that, the descendants of spatial transcriptomic or proteomics look good to me.

[brianraymor]

I can decompose the spatial cases into individual supported assays if that's preferable.

[jahilton]

The problem is that 10x multiome is a descendant of scATAC-seq.

👍

I don't think further decomposing is needed. I think adding a review step whenever EFO is updated should be sufficient. And establishing a 'supported assay' list will also help as it will narrow the scope of that review

[jahilton]

With the updated submission policy, our only smFISH Datasets (which are private) will be removed, and no more will be accepted. So the smFISH and its descendants row can be simplified to MERFISH EFO:0008992

[brianraymor]

Updated. Added a note to Update requirements for suspension_type.

[brianraymor]

Per conversation with @jahilton and @BAevermann - it does not currently make sense to allow "epigenomics" as a value for mCT-seq. It is unsupported by the updated submission policy. There are no published datasets with this assay+modality combination. It's ~struck~ above.

[brianraymor]

Based on the renewed discovery for 10X multiome, I'm reverting this issue from schema 5.2.0 and re-opening.