A London biotech with five employees and a Y Combinator pedigree has made a claim that would give most oncologists whiplash: it can "cure all cancers" by targeting DNA directly, unlocking roughly 80% of disease drivers that have eluded drugmakers for decades. The assertion is bold. The mechanism, intriguing. The supporting data? Nonexistent.
FinalDose, founded in 2025 and part of Y Combinator's Spring 2026 batch, is pitching what it calls a "programmable DNA drug"—a molecule that scans the genome, recognizes a specific sequence, and flips a molecular kill switch. The premise rests on a familiar frustration in drug development: most cancer drivers hide in what's known as the "undruggable proteome," that stubborn 80 to 85 percent of human genes that lack the convenient binding pockets small molecules need to latch onto. Can't drug the protein? Drug the DNA instead.
It's seductive, especially now. The broader field of programmable medicine is finally gaining clinical traction. The first CRISPR therapy won FDA approval in December 2023. DNA-drug conjugates capable of performing Boolean logic were published in Nature Biotechnology in March 2026. A separate Nature paper in May demonstrated that Cas12a2 enzymes can be programmed to kill human cells via RNA-triggered DNA damage. The infrastructure for sequence-level intervention is maturing, in other words. The question is whether FinalDose's particular approach can navigate the formidable gaps between concept and clinic—and whether anyone should take a "cure all cancers" pitch seriously before seeing a single data point.
The 80 Percent Problem
The "undruggable" label traces back to a 2002 Nature Reviews Drug Discovery estimate by Hopkins and Groom, who calculated that the druggable human genome sits at roughly 10 to 15 percent—genes encoding proteins with binding pockets that small molecules can exploit. That left oncogenes like KRAS, MYC, and a host of transcription factors tantalizingly out of reach. Over the past two decades, the industry has chipped away at the margins. PROTACs, molecular glues, covalent inhibitors—all have nudged some "undruggable" targets into the clinic. But the vast majority remain inaccessible.
FinalDose's proposition? Bypass proteins entirely. According to the company's YC launch post from May 12, the platform uses a single "smart" molecule that searches the genome, binds to a specific DNA sequence via Watson-Crick pairing, undergoes a conformational change, and triggers cell death. Swap the guide sequence, the company claims, and you have a new drug. "Same molecule, new guide → new drug."
The founders bring relevant credentials, at least on paper. Jeff Liu, the CEO, holds an Oxford PhD in oncology and previously raised funding across two rounds for Vivid Dx, a diagnostics startup—a $5 million seed and a $10 million pre-Series A. Li-Yao Huang, the chief scientific officer, has an Oxford DPhil in biochemistry. Steven Lin, the CTO, earned his PhD in computational biology, also at Oxford. But as of late May, no peer-reviewed publications, patents, or independent validation of the platform have surfaced. The company's LinkedIn feed—627 followers—emphasizes the "streetlight effect" critique: that researchers have been looking where the light is good (proteins) rather than where the problem actually is (DNA).
Fair point. Also a marketing pitch awaiting proof.
A Crowded, Chaotic Inflection Point
FinalDose is hardly alone in chasing programmable, logic-driven therapeutics. The broader DNA nanotechnology market was valued at $4.51 billion in 2024 and is projected to grow at a 19.9 percent annual rate through 2030, according to Grand View Research. NovaOne Advisor pegs 2025 sales at $6.15 billion, climbing to $37.86 billion by 2035. The numbers vary depending on who's counting, but the signal is clear: investors believe DNA-based devices have legs.
In March, a team led by Nicolas Winssinger at the University of Geneva published DNA-drug conjugates that perform AND-gate logic across two cell-surface biomarkers, triggering a hybridization chain reaction to amplify intracellular drug delivery. The system, described in Nature Biotechnology, doesn't target genomic DNA but demonstrates that DNA modules can "compute" at the cellular level—sensing, deciding, acting. It's an existence proof for programmable delivery.
Around the same time, researchers showed that Cas12a2, an RNA-activated nuclease, can be programmed to indiscriminately shred double-stranded DNA inside human cells, triggering apoptosis. The work, published in Nature in May after acceptance in late March, sketches a pathway to sequence-specific cell killing. Whether it can be harnessed safely—without collateral genomic damage—remains unclear.
Elsewhere, Senti Biosciences announced in April a Cell Systems publication on NOT-gated CAR-T designs, adding safety logic to engineered immune cells. Autonomous Therapeutics is developing "disease-activated RNA medicines" that translate only in cells meeting specific criteria, partnering with MiNK Therapeutics to target metastatic tumors. DNA Nanobots, a U.S. startup that closed a pre-seed round in January 2024, is building functionalized DNA nanoparticles for payload delivery.
None have reached the clinic yet. But the convergence is palpable, and perhaps more rapid than the field anticipated.
The Delivery Dilemma
If FinalDose's molecule needs to reach genomic DNA inside the nucleus to recognize and bind its target sequence, it faces one of biotech's most stubborn barriers: delivery. Macromolecules struggle to cross cell membranes, escape endosomes, and traverse the nuclear envelope. Reviews published between 2023 and 2026 catalog strategies—cell-penetrating peptides, nuclear localization signals, lipid nanoparticles—but none offer a silver bullet for large, complex DNA-binding constructs.
DNA origami and nanostructures, often touted for their programmability, bring additional headaches. A 2024 analysis in Nucleic Acid Insights highlighted stability, immunogenicity, and GMP manufacturing as persistent gaps between preclinical promise and clinical application. If FinalDose's platform relies on similar architectures, those challenges apply in full.
Then there's specificity. Genotoxic agents that damage DNA indiscriminately have narrow therapeutic windows and brutal side effects. FinalDose's Watson-Crick pairing premise suggests high sequence specificity, but genomic DNA is long, repetitive, and structurally complex. Off-target binding—especially in chromatin-dense or homologous regions—could trigger unintended cytotoxicity. The FDA's January 2024 guidance on genome editing safety sets expectations for next-generation sequencing and orthogonal off-target assessment, even for products that don't technically "edit" but engage the genome. Those standards will likely apply here too.
Regulatory Tailwinds, Regulatory Headwinds
FinalDose may benefit from evolving regulatory frameworks designed to accommodate individualized and platform-based therapies. In February, the FDA released draft guidance on a "Plausible Mechanism" framework for therapies targeting patient-specific mutations—situations where randomized controlled trials are infeasible. If FinalDose's guides can be swapped modularly to address different oncogenic sequences, the guidance could streamline approval for families of variants under a single platform, provided the mechanistic anchor and CMC controls are robust.
Project Optimus, the FDA's ongoing initiative to reform oncology dose-finding toward optimal biological dosing rather than maximum tolerated dose, is another variable. A programmable cytotoxic agent with a binary kill mechanism will face scrutiny on therapeutic index and translational pharmacokinetics early. The bar for demonstrating a meaningful safety window has risen considerably.
The first CRISPR therapy's approval in December 2023 for sickle cell disease—Casgevy—signaled that regulators are willing to greenlight programmable genomic medicines when the data are strong. But the jump from ex vivo gene editing of patient cells to systemic delivery of a DNA-targeting cytotoxic agent? Substantial doesn't quite cover it.
The Gap Between Claim and Clinic
Oncology remains the industry's largest growth engine. Sales hit an estimated $288 billion in 2025, with innovation concentrated in targeted modalities. Antibody-drug conjugates alone reached $18.8 billion in sales in 2025 and are projected to hit $36 billion by 2030, according to IQVIA. The appetite for precision oncology tools is real, and growing.
But FinalDose's "cure all cancers" positioning sits uncomfortably—awkwardly, even—with the lack of published data. The company's YC page and LinkedIn posts argue that most disease starts in DNA, that historical protein bias has left 80 percent of targets unaddressed, and that sequence-level specificity via Watson-Crick pairing solves the selectivity problem. These are hypotheses. Not evidence. No in vitro kill curves, no in vivo efficacy, no toxicology, no delivery method validation. Nothing has been disclosed publicly.
The founders' backgrounds offer some reassurance. Liu's experience raising capital and navigating diagnostics development at Vivid Dx (which uses Raman spectroscopy and AI for rapid sepsis detection) suggests operational competence. Huang's biochemistry credentials and Lin's computational biology expertise are relevant. But diagnostics and therapeutics are different beasts, and DNA-targeting cytotoxic drugs carry risks that lateral-flow assays simply do not.
What Comes Next
For FinalDose to move from YC Demo Day talking point to credible therapeutic platform, it will need to demonstrate several things, and quickly: proof of concept in cells, evidence of sequence-specific binding and killing without off-target toxicity, a viable delivery strategy (systemic or otherwise), and preclinical safety data in animals. The company will also need to clarify its location—YC lists London, while LinkedIn says San Francisco—and close a seed or Series A to fund those experiments. As of late May, no funding announcement has surfaced.
The broader programmable medicine landscape suggests the scientific infrastructure is catching up to the ambition. DNA-drug conjugates with logic gates, RNA-triggered cell-killing systems, logic-gated CAR-T cells—all are moving from academic proof-of-concept to translational exploration. Manufacturing, delivery, and regulatory pathways are crystallizing. If the fundamentals hold, the 2030s could see a wave of DNA-level interventions reaching patients.
But the distance between a compelling pitch and a validated therapeutic is measured in years, dollars, and failed experiments. Lots of failed experiments.
FinalDose has articulated a vision—one that resonates with a real industry pain point and arrives at a moment of genuine technical convergence. Whether the company can execute on that vision, or whether it becomes another cautionary tale of overreach, depends on data it has yet to produce. For now, the 80 percent remains stubbornly undruggable. And the claim to "cure all cancers" remains exactly that: a claim, hanging in the air, waiting for the science to catch up.
