chanzuckerberg / single-cell-explorer

Hosted version of cellxgene
MIT License
11 stars 2 forks source link

Side by Side Embedding Feature #922

Open niknak33 opened 4 months ago

niknak33 commented 4 months ago

Multimodal exploration of spatial data with side-by-side embeddings

Rationale:

This is foundational work to support data consumers’ ability to generate hypotheses from spatial data. Being able to compare the location of cells in physical space with their position in UMAP or other lower-dimensional “transcriptomic space” enables data consumers to investigate how gene expression profiles change in different cell types and tissue locations.

User stories:

As a user, I want to directly compare cells’ placement in physical space (spatial embedding) with its representation in a low-dimensional embedding (e.g., UMAP or tSNE) so that I can better understand how patterns in gene expression vary across space.

Requirements:

Design - https://www.figma.com/file/eJ3acPxG04PKVlom6eS8ft/Spatial?type=design&node-id=304%3A195512&mode=dev @hthomas-czi

Sub-Tasks:

tihuan commented 4 months ago

@seve I don't remember if the data point limit we currently support is memory bound or CPU/GPU bound, but maybe worth creating a ticket to double check that the side by side mode will work with the max data points we support 😊?

Thank you!

tihuan commented 4 months ago

bumped up to P0, since this is required for launch