One disease map, three engines.
Juva turns disease biology into ranked candidates, then lets human cells judge them. Here is what each engine does and where it stands today.
Footage: liquid-handling robot, illustrative
Therapy Design AI
Ranks combinations of approved drugs against a disease signature.
Juva reads the gene-expression signature of a disease, maps the pathways it depends on, and searches combinations of approved drugs that hit several of those pathways at once. Approved drugs carry years of safety data, so a good combination has a shorter road to the clinic than a new molecule.
- Decomposesplit the signature into pathway dependencies
- Searchcombinations of approved drugs that cover them
- Rankpredicted cancer-cell death, healthy-cell safety, known dosing
Footage: automated dosing into a well plate, illustrative
Vaccine Design AI
Designs peptide and mRNA vaccine candidates for cancer and aging targets.
The same disease map points at proteins the immune system could learn to hunt. Juva scores each peptide candidate on four axes before anyone synthesizes it, then drops the ones that look too much like your own tissue.
- Immune responsepredicted MHC binding and T-cell activation
- Selectivitydistance from the human self-peptidome
- Stabilityhalf-life and aggregation risk
- Manufacturabilitysolid-phase synthesis yield
Footage: DNA model, rendered, illustrative
Validation Loop
Tests predictions on cancer cells and healthy cells side by side.
A candidate that kills cancer cells and healthy cells alike is poison. Juva will run every top-ranked candidate on both in the same assay, record the gap between them, and feed the readout back into the model before the next batch.
- Cancer armcell death over seven days
- Healthy armthe same compounds on healthy human cells
- Feedbackeach readout updates the ranking model
Footage: microscope objective turret, illustrative
Why combinations of approved drugs.
More pathways covered
Tumors and aging tissue lean on several pathways at once. Two or three drugs can block routes one drug leaves open.
Known safety
Approved drugs arrive with dosing and side-effect records, which shortens the road to a trial.
Harder to escape
Cells that adapt around one drug meet a second block. Resistance takes longer to appear.
The loop.
Read
Map a disease from its gene-expression signature.
Which genes run hot or cold in sick tissue, and which pathways those genes feed.
Rank
Score thousands of drug combinations in silico.
Coverage of the pathways, predicted selectivity, and dosing we already know is safe.
Test
Put the top candidates on human cells.
Cancer cells and healthy cells, same compounds, same conditions, side by side.
Learn
Advance the winners. Retrain on every result.
Failures carry as much signal as hits. Each round makes the next ranking sharper.
Public data we plan to build on first.
Open datasets let anyone check the first rankings. Our own lab data joins once the first assays run.
- DepMap
- Genetic dependencies measured across more than a thousand cancer cell lines.
- GDSC
- Drug sensitivity screens that link compounds to cell-line response.
- LINCS L1000
- Gene-expression changes after cells meet thousands of compounds.
- GTEx
- Expression in healthy human tissues, the baseline for selectivity.