AI-programmable DNA switches

Therapy that switches on only inside cancer.

OncoSwitch designs short pieces of DNA that stay silent in healthy cells and turn a gene therapy on inside tumor cells.

See how a switch works Partner with us
StagePreclinical
First targetLiver cancer (HCC)
FundingPre-seed closed
Featured in Science Times, October 2026 Co-founded with UCSF professor Nadav Ahituv Physics-informed AI plus our own wet-lab data R&D in MBC BioLabs

Inside the therapy

Every gene therapy carries DNA instructions.

The switch

A switch is a short stretch of that DNA designed by AI.

Inside the liver

In tumor cells, the switch turns the therapy on.

Everywhere else

In healthy cells, it stays off.

01The problem

Today, gene therapy switches on wherever it lands.

A therapy carried into the liver reaches healthy cells as well as the tumor. If it switches on in both, healthy tissue takes the hit. That is where many side effects come from.

Move the scanner over the tissue. Purple means the therapy is switched on.

Tumor cellsOn
Healthy cellsOff
Hits healthy tissueNo

With a switch in front of the therapy gene, healthy liver cells keep it silent. Only the tumor lights up.

02How a switch works

A short piece of DNA, placed in front of the therapy gene.

Tumor cells have distinct patterns of protein activity. A switch is designed to respond to these patterns, activating only in tumor cells.

SwitchOn

In tumor cells, the right combination of regulatory proteins binds to the DNA switch, activating the therapeutic gene and producing the therapy.

Tap a part of the DNASwitch

Short stretches of DNA letters that tumor-specific proteins recognize. Our AI chooses which sites to use and how to arrange them.

03The platform

AI designs. The lab tests thousands at once. The model learns.

A massively parallel reporter assay (MPRA) gives every design its own barcode, so thousands of candidate switches are measured in tumor and healthy cells in a single experiment. Every result trains the next round of designs.

01 DesignAI writes new candidate switches
02 TestMPRA measures every design at once
03 LearnThe best designs teach the model

A simplified simulation of 10 design-test-learn cycles. The actual number of cycles varies depending on the experimental goals and conditions.

Cycle00 / 10
PhaseReady
How designs score
On everywhereOn only in tumor
Best switch so far

04Fits existing therapies

One switch. Any delivery system.

The switch sits inside the therapy's DNA, so it works with the carriers developers already use. The payload itself does not change.

Drag the capsule onto a carrier, or tap one.
Installed inNone yet

05Validation pathway

Proven level by level.

Every switch starts as a computer design, then has to clear increasingly realistic biology before it reaches a partner's therapy.

In progress
Level 01, cell lines

High-throughput screening measures switch activity and selectivity across cancer and non-cancer cell lines, starting with liver cancer models.

06Roadmap, 2026 to 2028

What we have built, where we are, and what comes next.

Achieved so far
  • Pre-seed round closed
  • Computational prototype built
  • R&D running in MBC BioLabs
  • Team in place across AI, genomics and BD
ProgramAI designCell linesPrimary cellsOrganoidsIn vivoPartnering
Liver cancer (HCC) switchesLead program
Now: cell-line screening
Additional cancers2028 objective
Planned
  1. Phase 01PreparationTeam, platform, librariesDone
  2. Phase 02MPRA and AI modelsMPRA cycles, AI baseline, closed loopWe are here
  3. Phase 03MVP and BDMVP switches, assays, BD outreachNext
  4. Phase 04PilotProduction library, validation, pilotsPlanned
  5. Phase 05LicensingIP, licensing, Series A prepPlanned

Objectives by 2028

  1. Reproducible proof of concept for the switch platform
  2. Candidate switches validated across multiple cancers
  3. More commercial partnerships signed
  4. First proprietary therapeutic candidate selected

07Team

AI, genomics and drug development in one team.

The people building OncoSwitch, from UCSF functional genomics to two decades of pharma launches.

08Partner with us

Build safer gene therapies with us.

We work openly, share data as it comes in, and co-develop with the teams who will take therapies to patients.

Pharma and biotech

Bring your payload and target. Start with a paid pilot, move to collaboration, then license the switch.

Dexter Sugiono, Head of Business DevelopmentEmail Dexter Investors

Our pre-seed round has closed. Talk to us about the next stage of the platform.

Malika Gallyamova, Founder and CEOEmail Malika Research institutions

Co-develop and validate switches in new cancers and tissue types.

Vitalii Volkov, Co-founder and CTOEmail Vitalii
General inquiriesconnect@oncoswitch.ai
How partners payPaid pilotsPlatform licensesMilestonesRoyalties