Building the virtual molecular layer for tissue biology¶

Aurora predicts spatial gene expression from routine H&E images, helping researchers explore molecular patterns across samples, cohorts and disease.

The section at low magnification, pink and purple tissue on a pale mount, with three rings marking the places the worked example shows close up. — An open-access slide of a melanoma metastasis in a lymph node. Both panels are the same region of it. The right one maps `MS4A1` over the 24,308 retained tissue tiles in that region. Inspect the slide and both gene maps.

MS4A1 predicted across every retained tissue tile inside that same frame, brighter where the predicted value is higher. — An open-access slide of a melanoma metastasis in a lymph node. Both panels are the same region of it. The right one maps `MS4A1` over the 24,308 retained tissue tiles in that region. Inspect the slide and both gene maps.

What Aurora helps you do¶

Past tissue. Future questions. A cohort assembled years ago can meet a question that emerged yesterday.
Find patterns at cohort scale. Explore virtual molecular variation across samples and biological groups.
Prioritise what to pursue. Focus validation and experimental effort on the signals, samples and hypotheses that matter most.

Explore research applications on the website, or explore one real slide here.

From existing tissue to a virtual molecular view¶

Virtual spatial transcriptomics is predicted, spatially resolved gene expression from H&E images. Aurora returns it on the slide's own coordinates as one transcriptome-wide matrix. No spatial assay is run.

Choose the workflow that fits your research¶

The tissue is already here. A new layer of biology is within reach.

Upload your slides

The simple path

Aurora Direct takes an H&E image and an email address, with no code and no account. Submit images for analysis. From code, install and prepare a slide, or compare the two routes.
Keep your slides with you

When images must stay in your environment

Run DeepSpot-H on your own machine and send the numbers it produces. Run DeepSpot-H locally.

Analysis is for academic and non-profit research. Commercial evaluation and use are governed by a written agreement with Aurora. Discuss commercial access.

The models behind a prediction¶

Two models make each prediction. DeepSpot-H reads the tissue's structure in each part of the slide, so DeepSpot-M can predict spatial gene expression there. Learn about DeepSpot-H or learn about DeepSpot-M.