Find new research directions in your existing cohorts.
An existing cohort can become the starting point for a new biomarker programme. A finding in a few profiled cases may raise a question about the wider collection, or a completed study may become relevant to a target that emerged later.
Aurora predicts spatial gene expression from H&E slides, allowing you to investigate those connections across samples and within their tissue. Earlier research can then contribute to new target hypotheses and questions about variation within an indication.
Applications
Develop new directions from existing research
- Explore
Revisit a completed study with a new question
When a new target enters a programme, earlier cohorts offer material in which to investigate it. Predicted gene maps let you return to those cases and examine its spatial pattern, even if the original study never measured the gene.
Built onOut-of-the-box predictionComparative spatial analysis
DeepSpot-M produced a virtual spatial atlas of 28,664 slides across 32 cancer types, published on Hugging Face.[1, 2]
- Explore
Follow unexpected connections across the whole cohort
A candidate biomarker in a profiled subset raises a wider question: does its predicted pattern also appear in the other cases? Comparing the available slides lets you investigate that possibility and develop research hypotheses about differences within the cohort.
- Develop hypotheses about subgroups of patient samples for research
- Include existing sample-matched bulk RNA in the spatial prediction
Built onOut-of-the-box predictionComparative spatial analysisSample-matched bulk RNAThe analysis report and the explorer
The web upload does not take a bulk profile. It goes in through the Python package and the API, on the Enterprise route.Learn about Enterprise
DeepSpot-M predicts 19,338 protein-coding genes at every retained tile of a slide, from the H&E image alone.[1]
- Extend
Take a study beyond the panel it was designed around
A measured target may prompt questions about related genes or neighbouring tissue that the panel did not cover. With our team, you can explore additional predicted genes in those same regions or revisit selected genes lost during quality control, giving the finding a broader molecular context.
Built onPanel extensionFinetuneRestore genes
Arranged with our team, on the Enterprise route.Learn about Enterprise
The DeepSpot-M preprint demonstrates transfer to held-out cancers, and adaptation to a new cohort from one slide.[1]
- Look ahead
Begin each new study with a wider view
Predicted maps let you work backwards from the biological comparison a study should make to the samples and regions it should include. That gives panel design and capture placement a shared rationale before the experiment begins.
Your route
Discuss the next direction for your programme
Bring a cohort, a biomarker question or a completed study to the conversation.
Companies running translational and discovery programmes.
Worked examples
Explore how existing tissue supports new research

Give your archive a new role in discovery.
Bring unprofiled cases into molecular research and look for shared patterns across your archive, using the slides you already hold.
- You bring
- A digitised H&E slide for each sample you want to explore.

Bring a spatial perspective to your molecular data.
Revisit existing bulk RNA findings in their tissue context, using sample-matched profiles to inform predicted gene maps.
- You bring
- A digitised H&E slide and a bulk RNA table from the same tissue block.

Shape your next experiment with a wider view.
Build your study around the patterns you want to investigate, with a view of candidate samples and tissue regions before the assay begins.
- You bring
- The slides you are choosing between.
References
- [1]Nonchev K, Dawo S, Silina K, Koelzer VH, Rätsch G. DeepSpot-M: a multimodal foundation model for transcriptome-wide virtual spatial transcriptomics from histology. medRxiv. Preprint, 2026. https://doi.org/10.64898/2026.06.19.26356060 (opens in a new tab)
- [2]Nonchev K, Dawo S, Silina K, Koelzer VH, Rätsch G. TCGA Virtual Spatial Transcriptomics Atlas. Hugging Face. Dataset, 2026. https://huggingface.co/datasets/ratschlab/TCGA_virtual_spatial_transcriptomics_atlas (opens in a new tab)