A navigable, evidence-graded map of disease biology. 34,000 diseases. Resolved to cell type.
Target discovery brings several evidence layers together: genetics, single-cell expression, perturbation, network biology, structure. Each answers a different question, and integrating them with cell-type resolution carried through to the ranking is a methodological challenge the field has been refining for years.
Disease Atlas by Euretos is now live. It is a navigable, evidence-graded map of disease biology across 34,000 diseases, resolved to cell type, built from primary biological data rather than from papers about that data.
For each of 34,000 human diseases, Disease Atlas maps the biology from organ to tissue to cell type. Pick a disease and you can see which cell types are involved, which targets sit causally upstream of the phenotype, what the safety landscape looks like, which therapeutic modalities are open, and what has already been tried.
The foundation underneath all of this is 3,500+ cell-type expression profiles from integrated single-cell data, harmonised across organ and tissue ontologies, joined with genetics, perturbation responses, molecular interactions, and curated patient evidence. Every entry traces back to a primary source.
Both target discovery and target assessment work off the same map. A researcher can move from a first observation to a conclusion they can stand behind without switching tools or rebuilding context.
The Target Prioritisation Score ranks the full protein-coding genome against each of those 34,000 diseases. Five complementary evidence layers contribute, each answering a different question about cause:
All five layers run on quantitative biological inputs. None of them rank a gene by how often it has been mentioned in disease literature. Candidates emerge because the biology converges, independent of literature mentions.
For a known target, the same framework runs in reverse. Which diseases is this target causally implicated in? Where does the cell-type biology point? The result is a ranked list of candidate indications resolved to the cells where the target actually operates.
Most platforms display cell-type expression as evidence on a target page, treated as a filter or a tag. Disease Atlas ranks genes per cell type per disease. The cell type is a ranking axis, not an annotation.
This changes which candidates appear at the top. In inflammatory bowel disease, three of the top-ranked candidates surface only when the ranking is resolved to the cell types where the disease operates. Bulk-disease rankings miss them. The biology is the same; the resolution is what changes the answer.
For any gene-disease pair, Disease Atlas surfaces a complete and auditable rationale. Six dimensions sit side by side: causal evidence, the mechanistic cascade from gene to pathway to cellular function to tissue function to phenotype, expression context, network position, tractability across modalities, and safety. Each dimension is resolved to the cell types where the disease operates. Every score traces back to a primary source.
The rationale is the artefact the researcher takes into a project committee. It holds up to questions because every layer is structured the same way and every piece of evidence is named.
We made a few choices deliberately, and these are where the platform departs from what is currently available.
Primary biological data, not extracted claims. Disease Atlas runs Mendelian randomisation on GWAS effect sizes. It processes single-cell expression matrices, computes perturbation magnitudes, and integrates structure data. It is not built by extracting “causal” statements from paper text. The reproducibility difficulties that have shadowed published claims and conclusions sit downstream of the layer the platform works from.
Quantitative causal inference, not citation counting. The five evidence layers are integrated with explicit causal weighting. Each layer answers a different question about whether perturbing a gene changes the disease phenotype, rather than a question about how often the gene appears in the same paper as the disease.
Cell-type-axis ranking, not cell-type evidence display. Cell type drives the ranking itself. Disease Atlas resolves to 3,500+ cell types, where reference platforms typically work with a few hundred.
Calibrated probabilities, not one number. Each pair carries three calibrated probabilities, one for each step in the translational journey. The first reflects how strongly the biology supports cause. The second reflects how likely the target is to reach clinical trials. The third reflects how likely it is to be approved. When a candidate scores high on biology but low on the chance of reaching the clinic, it is a target the field has not yet developed. Uncertainty is broken out by source. Some comes from ambiguity in the data itself. Some comes from layers that disagree. Some comes from training that does not hold steady across folds. Each call also carries a confidence flag tied to the disease itself. The flag tells you how much data the platform has on that disease. A rare disease with thin evidence behind it shows up as a low-confidence call. The flag is there in plain view. Calibration is measured on held-out data and reported, not asserted.
Deterministic by design. No large language model sits in the reasoning path. Every score is reproducible. Every claim traces to a tier-graded source.
Disease and target in one workspace. Disease biology and target ranking share the same cell-type scaffold. The detail page brings biology, tractability, safety, modality, and competitive landscape into one view per disease, rather than scattering them across separate tools.
The intent behind Disease Atlas is straightforward. Give the translational researcher a way to move from a first observation to a conclusion they can stand behind, without rebuilding the integration every time. Questions that previously required a dedicated project (pull the GWAS, run the MR, layer the single-cell, query the network, write the rationale) can now be explored, tested, and answered independently.
That is the work. The platform is the tool the researcher uses to do it.
Disease Atlas is live at ask.euretos.com. For readers who want the full methodology and the calibration diagnostics behind every claim above, the Disease Atlas Methodology and Validation paper is available in our methodology section. Over the next few weeks we will walk through the methodology in shorter form here. First, what we mean when we say “causal evidence.” Then, each of the five evidence layers and the question each one answers.
Follow along, and bring your hardest disease.