We use AI to cure cancer.
Palo Alto Molecules designs new medicines in weeks, not years, for targets the industry has validated but never been able to drug.
Every drug is a key for a lock.
The lock is a protein that drives a disease. The key is a molecule that fits that protein and nothing else. Biology has found thousands of locks. Making the key still takes years and tens of millions of dollars per attempt, and most attempts fail.
We are building the fastest key-maker in the world. Our mission is to make drug discovery so fast that curing cancer becomes a question of time, not luck.
A recursive loop.
Design
Our proprietary model designs a candidate molecule against the target.
Make
The molecule is synthesized.
Test
It is tested against human receptors in the lab.
Learn
Every result, positive or negative, goes back into the model, which designs the next version.
Each cycle takes weeks. Each cycle makes the model better.
Known locks. Missing keys.
We do not hunt for new targets. We work on targets that decades of research have already validated, where the obstacle is chemistry: nobody has made a molecule that hits the target cleanly and leaves everything else alone.
Selectivity is where traditional discovery stalls. It is what our model is built for.
Three programs. Weeks per cycle. A fraction of the cost.
Every program runs on the same design, make, test, learn loop. "Our spend" is total spend to date on that program. "Industry" is the published average cost and time for a traditional pharma program to reach the same stage.
| Program | Target | Indication | Where we are | Our spend | Industry cost to reach this stage | Industry time to reach this stage |
|---|---|---|---|---|---|---|
| RC01 / RC21 | Selective CB2 agonist | Neuroinflammation, Alzheimer's disease | In vitro validatedDesigned, synthesized, and selectivity confirmed in human receptor assays. Next: in vivo. | Under $30K | ~$3.5M | ~2.5 years |
| PAM-02 | Undisclosed | Inflammatory bowel disease, ocular inflammation | Design | Compute only | ~$1M | ~1 year |
| PAM-03 | Selective PARP1 inhibitor | Oncology | Design | Compute only | ~$1M | ~1 year |
- Design
- the model designs a molecule for the target.
- Synthesis
- the molecule is physically made.
- In vitro
- tested against human receptors in the lab, outside a living organism.
- In vivo
- tested in animals for effect and safety.
- IND
- the regulatory package that allows the first human trial.
- Phase 1
- safety in a small group of people.
- Phase 2
- does it work, in patients.
- Phase 3
- does it work better than the standard of care, at scale.
Industry benchmarks from Paul et al., Nature Reviews Drug Discovery, 2010, Table 1. Out-of-pocket cost and time per program: target-to-hit $1M and 1 year; hit-to-lead $2.5M and 1.5 years; lead optimization $10M and 2 years. Excludes failed programs and cost of capital. Full discovery through a preclinical candidate: $13.5M and 4.5 years.
Data built for the model.
Most of the industry's experimental data was collected over decades by different teams, in different formats, for different purposes. It cannot train a model.
Ours is generated for the model from the first experiment: structured, consistent, and verified in the lab. Every molecule we make teaches the next one.
Peter Zhang
Co-founder and CEOStanford, Computer Science and Mathematics. Computational researcher in the laboratory of Brian Kobilka, 2012 Nobel Prize laureate in Chemistry. Designed RC01, the first molecule out of the loop to be validated against human receptors.
Michael D. Sacco, PhD
Scientific AdvisorPostdoctoral researcher in the Kobilka laboratory at Stanford. Leads experimental validation.
You validated the target. We make the molecule.
Palo Alto Molecules delivers assay-validated candidates against hard-to-drug targets and partners with pharmaceutical companies to take them into the clinic.