Find the cells driving your disease. Before the lab.
Relation Therapeutics trains foundation models on single-cell genomics to surface the cell populations behind a disease state, giving your discovery team target hypotheses grounded in human biology.
More than 90 percent of drug candidates entering clinical trials fail. A recurring factor: target hypotheses built from bulk tissue data that masks cell-type specificity.
Bulk RNA sequencing averages millions of cells into a single signal. Disease-driving populations, often a small fraction of total tissue, are invisible at that resolution. Programs advance on targets that look good in the average but behave differently in the cell subtype that actually drives pathology.
Single-cell genomics changed what is measurable. Foundation models trained on those measurements change what is computable. The tools your team needs to act on that shift are what we are building.
What the platform does
Trained on millions of single-cell RNA-seq profiles across human tissues and disease states. The model learns how cell types vary in expression, state, and behaviour across biological contexts.
Given a disease indication and tissue context, the platform identifies and ranks cell populations associated with the disease phenotype, grounded in human single-cell reference data.
For each identified cell population, the platform returns gene candidates ranked by expression specificity across tissues. Output is structured for direct use by bioinformatics and discovery teams.
How it works
Describe the disease phenotype, the tissue of interest, and any additional biological context. The platform accepts structured queries for both common and rare disease areas.
The encoder maps your query against its learned cell population space, drawing on single-cell profiles from public atlases and curated disease datasets to find relevant cell types and states.
Receive a structured output: ranked cell populations driving the phenotype, and for each population a ranked list of target gene candidates with expression specificity across tissues.
Built on the largest public single-cell references
Our foundation model learns cell biology from the largest publicly available single-cell datasets. The Human Cell Atlas, CELLxGENE, and other curated public collections provide the biological ground truth the model generalises from.
Transformer architectures that represent individual cells as tokens enable the model to capture gene co-expression patterns, cell-state transitions, and cross-tissue variation at a resolution bulk sequencing cannot match.
Built by genomics researchers
Background in computational biology and genomics data systems. Spent years building research infrastructure for translational and academic programmes. Founded Relation Therapeutics in 2023 after seeing how often discovery programmes fail at the biology step, not the chemistry. Angel-funded February 2026.
Single-cell genomics specialist with years building cell-type annotation and transcriptomic analysis pipelines. Expertise in disease cell population identification and genomics data interpretation. Co-built the scientific core of the Relation Therapeutics platform from the ground up.
Work with us on your next target ID programme
Research partnerships and early access collaborations welcome. We are looking for discovery teams serious about cell-type grounded target selection.