What you get back¶
Your Aurora result is a spatial molecular view of tissue: predicted gene expression aligned to your H&E slide. Explore it in Aurora Intelligence or download it as AnnData for your own analysis.
What you can explore¶
- Genes across tissue. Find regions with higher or lower predicted expression.
- Regional patterns. Read molecular variation alongside tissue morphology.
- Differences between samples. Compare the same genes across a cohort, using consistent preparation and model settings.
- Questions to pursue. Choose promising regions, samples and genes for measured follow-up.
An example slide shows how these views connect to recognisable structures in real tissue.
Open it in Aurora Intelligence¶
Explore the result as a map, with genes placed on your slide's coordinates. In Aurora Intelligence, import a finished result from your inbox into a workspace.
When submitting from Python, set land_in_workspace=True to offer the result
to that inbox. Workspace privacy
explains who can access it after import.
What the values mean¶
Each value is predicted expression for one gene in one tissue tile. Compare predicted patterns with measured expression from follow-up assays. Use patterns to form hypotheses and focus experiments.
A gene without measured training data has a blank value (NaN), rather than
zero expression. Its measured_in_training field is False.
The model card states the expression scale under output.transform.
Analyse it in your own tools¶
Every result is an .h5ad file: one row per retained tile and one column per
gene. Spatial coordinates and model information travel with the values.
The result file reference covers AnnData fields, gene identifiers and resolution metadata.
Find a gene¶
Use the gene symbols in feature_name to browse the result. For joins between
datasets, use the Ensembl identifiers that index the gene columns.
Gene lookup explains aliases and API queries.