Wet-lab guide#
This guide is written for wet-lab researchers and collaborative users who may not want to start from API names or implementation details.
What goes in#
TurboCell Atlas needs three practical ingredients:
a reference atlas represented as embeddings plus metadata
a query that represents the cell state you care about
metadata fields that let you judge whether the retrieved cells make biological sense
What comes out#
The package does not return only one score. The useful output is a bundle:
ranked cells or nearest-neighbor tables
metadata columns that tell you what those cells are
scenario or benchmark summaries that help you compare exact and compressed retrieval
figures that make it easier to discuss the result with collaborators
How to read the output#
Read the output in this order:
Check whether the returned cells look biologically coherent.
Check whether metadata filters changed the answer in an expected way.
Compare compressed retrieval with exact retrieval if the workflow includes both.
Only then move to a stronger interpretation.
Why this matters#
The package is most useful when a ranked result can be explained in ordinary biological language. If the retrieval table cannot be connected back to disease context, tissue, sample, or cell-type annotations, the result is much harder to trust.