Twenty-four years after scientists finished sequencing the human genome, here's what we're still stuck with: roughly 85 to 90 percent of the proteins that actually drive cancer remain beyond the reach of any drug.
That figure—first etched into the field's collective consciousness by Hopkins and Groom in a 2002 Nature Reviews Drug Discovery paper—has proven maddeningly durable, though it's worth noting it's a historical estimate rather than a current measurement. PROTACs came along. Molecular glues arrived. The "druggable proteome" expanded, but only at the margins. MYC still can't be touched. TP53 remains elusive. APC, a major player in colorectal cancer, has no drug with its name on it. These proteins lack the deep pockets that small molecules need to latch onto, and biologics can't easily reach the cellular interiors where the damage happens.
A small cohort of biotechs thinks the problem isn't the proteins themselves—it's the fact that we've been aiming at the wrong layer entirely.
Instead of chasing proteins after they're manufactured, these companies are building therapeutics that read DNA directly. The pitch sounds almost too tidy: program a molecule to recognize a specific genetic sequence inside a tumor cell, trigger a kill switch, and suddenly you've got access to targets that traditional drug development has written off as impossible. Whether it actually works in patients is another question. But the underlying logic has begun attracting serious money and, perhaps more tellingly, cautious interest from regulators who've spent the past few years learning to evaluate increasingly exotic modalities.
Where the Old Math Breaks Down
The constraint isn't just theoretical. Some of oncology's most notorious drivers—MYC overexpression, TP53 mutations, APC loss—have swallowed billions in failed drug programs. Targeted protein degradation has expanded what's technically "druggable," but delivery problems and tissue distribution headaches persist.
DNA-addressable therapeutics flip the script. The core technology—triplex-forming oligonucleotides that bind to specific genomic sequences and induce localized DNA damage—has been kicking around preclinical labs for years. A 2021 Nature Biotechnology paper showed that TFOs targeting amplified HER2 gene loci could trigger copy-number-dependent cell death in mouse models, essentially turning gene amplification against itself. Yet those early systems couldn't get out of the lab; stability and delivery were deal-breakers.
What's shifted, according to researchers and company founders in the space, is a convergence that makes 2026 feel different. DNA nanotechnology has matured past academic proof-of-concept. Logic-gated delivery—borrowed from synthetic biology and now showing up in CAR-T designs—has demonstrated that you can program therapeutics to respond to multiple inputs, not just one. And regulators, having spent the better part of a decade wading through oligonucleotide and gene therapy submissions, seem more willing to engage with molecules that behave like software.
In March, researchers at the University of Geneva published work in Nature Biotechnology on DNA-drug conjugates that execute Boolean logic circuits. The system required two cell-surface biomarkers—EGFR plus PD-L1, or EGFR plus PTK7—to trigger a chain reaction delivering cytotoxic payloads. Signal amplification topped 100-fold. Cell-selective killing worked across multiple cancer lines. It's not a therapeutic yet, but it's the kind of proof point that makes investors lean forward.
A Nature Nanotechnology paper from late 2024 described a pH-responsive DNA "robotic switch" that exposes cytotoxic ligands only in acidic tumor microenvironments—exploiting tumor biology itself for selectivity. The FDA's June 2024 final guidance on clinical pharmacology for oligonucleotide therapeutics signaled that regulators are getting comfortable with DNA-like molecules circulating in the body, which should smooth the path for conjugates and nanostructures that behave similarly, though perhaps not as smoothly as developers hope.
The Convergence Thesis
Three forces are colliding, and the timing has biotech watchers and venture investors paying closer attention than they were even a year ago.
First, the technology stack is catching up. DNA origami and stimuli-responsive nanostructures have evolved from clever academic tricks into platforms with something approaching reproducible manufacturing. That doesn't mean scaling to GMP-grade material is trivial—it's not—but the gap between lab bench and production line is narrower than it used to be.
Second, logic gating is having a moment. A2 Biotherapeutics received FDA Fast Track designation in April for A2B543, a CAR-T therapy that combines tumor antigen recognition with a NOT-gate for HLA loss of heterozygosity, plus a membrane-tethered IL-12 booster. That's three layers of programming designed to ensure activation only in the right cells. Strand Therapeutics, which dosed its first patient with a programmable mRNA therapy (STX-001) in 2024 and presented Phase 1 data at ASCO last year, is layering similar logic into self-amplifying RNA circuits. The regulatory wins suggest the FDA is willing to evaluate multi-input, programmable systems—important precedent for DNA-reading drugs, assuming those drugs ever make it to the clinic.
Third, the market wants in. Oligonucleotide therapeutics alone are projected to reach somewhere between $17.7 billion and $29.6 billion by 2030, depending on which analyst you ask (MarketsandMarkets and Grand View Research bracket the range). Oncology remains the largest bet in biopharma, with the overall market forecast to hit $419.8 billion by 2030 according to Frost & Sullivan. Investors and pharma partners are hunting for modalities that can access the 80 to 90 percent of cancer biology current drugs can't.
Enter the New Guard
FinalDose, a London-based startup that came through Y Combinator's Spring batch, embodies the most ambitious version of this thesis. According to the company's profile on YC's directory, FinalDose is building "programmable smart molecules" that recognize hard-coded DNA sequences in diseased cells and trigger a kill switch. The same molecular chassis can theoretically be reprogrammed with a new guide sequence to target a different disease—similar in concept to how CRISPR's Cas9 protein can be redirected by swapping in a different guide RNA.
CEO Jeff Liu, who holds an Oxford PhD in oncology and previously co-founded Vivid Dx (which raised $15 million), has been vocal on LinkedIn about the limits of protein-targeting modalities. In posts from late May and early June, Liu argued that even optimized small molecules and biologics max out at 30 to 50 percent target coverage, whereas DNA-layer recognition could theoretically reach 90 to 95 percent. He also claimed the team went from concept to lab-level proof of a DNA-triggered fusion protein in 40 days. That's unverified and not peer-reviewed—worth noting—but it suggests the company is moving fast, or at least wants to project that it is.
FinalDose's team is small, five people as of early June, and the company remains in stealth mode with no public funding announcement beyond its YC participation. Demo Day is scheduled for June 16, which should clarify whether the company has lined up a seed round and what the near-term roadmap looks like. Until then, much of what FinalDose is building remains speculative.
Elsewhere in the ecosystem, Flagship Pioneering launched Serif Biomedicines in April with a $50 million initial commitment. Serif's platform centers on modified DNA medicines that promise non-viral, redosable, durable expression without genomic integration. The company has hinted at preclinical data in non-human primates showing tolerability and sustained expression after intravenous administration. If validated, that would address one of the major objections to DNA-based therapeutics: delivering DNA without viral vectors or integrating it into the genome has historically been inefficient and transient.
Strand Therapeutics is further down the clinical path. The company's STX-001, an intratumoral IL-12-encoding mRNA with programmable logic, has shown early safety and activity signals in Phase 1 trials for solid tumors. Strand raised a $153 million Series B in August 2025 and continues to advance programmable in vivo CAR-T concepts using circular RNA logic circuits.
What ties these companies together—FinalDose, Serif, Strand, A2 Bio—is the conviction that programmability, not just potency, is the next frontier. Traditional drug discovery optimizes binding affinity and selectivity one molecule at a time. Logic-gated and DNA-reading platforms aim to build universal chassis that can be reprogrammed for new indications by swapping in new guides or rewiring decision circuits.
Whether that vision holds up in the clinic is the open question, and the one that will ultimately determine whether this is a genuine shift or just another overhyped modality cycle.
The Regulatory Unknown
The regulatory pathway for DNA-addressable cytotoxic drugs is still being written in real time. Will they be reviewed as oligonucleotide therapeutics? Gene therapies? Small-molecule conjugates? The FDA's April draft guidance on using next-generation sequencing to assess genome editing safety suggests regulators are thinking hard about how to evaluate modalities that interact with genomic DNA. Even if a therapeutic doesn't edit the genome, sponsors may need to demonstrate that DNA sequence recognition is specific enough to avoid genotoxic off-target effects at therapeutic doses.
The 2024 final guidance on oligonucleotide clinical pharmacology—covering distribution, metabolism, and immunostimulation—offers a starting framework. But DNA-drug conjugates and logic-gated nanostructures will likely push the boundaries of existing categories, which means the first applicants may face longer review timelines and more granular questions about mechanism of action.
Manufacturing and quality control aren't trivial either. Academic labs can hand-assemble DNA origami structures and affibody-DNA conjugates. Scaling to GMP-grade material with consistent stoichiometry and batch-to-batch reproducibility is a different engineering problem entirely. Multiple reviews from the past couple of years on DNA nanomedicine note the gap between elegant proof-of-concept systems and products that can be characterized, stored, and administered like conventional biologics.
Then there's the biology itself, which has a way of humbling even the most elegant platforms. Tumor heterogeneity and mechanisms like extracellular DNA amplification—highlighted in a February Nature Reviews Drug Discovery article—could drive resistance even to DNA-targeted therapies if the logic is too simple or the drug can't adapt to clonal evolution. A DNA-reading molecule that works beautifully in a genetically uniform cell line may struggle in the mosaic landscape of a real tumor. Multiplexed targeting and adaptive reprogramming may be necessary, which adds layers of complexity to development and manufacturing.
What to Watch
Still, the momentum feels real, or at least real enough that serious players are placing bets. Precision oncology is trending toward biomarker-driven and AI-enabled trial designs, creating infrastructure that DNA-reading therapeutics would need to demonstrate clinical utility. ASCO's recent themes have emphasized real-world biomarker integration and computational pathways to patient stratification—exactly the kind of environment genotype-addressable drugs would need to thrive.
McKinsey's December 2025 analysis of oncology clinical development noted that 70 percent of U.S. counties lack an active cancer trial, underscoring the need for decentralized and lower-burden protocols. If DNA-programmable drugs can be designed with predictable pharmacology and straightforward patient selection criteria—sequence the tumor, match the guide—they could slot into leaner trial models. That's an if, not a given.
For investors and pharma watchers, the near-term milestones are clear enough. FinalDose's Demo Day on June 16 should reveal funding details and any technical specifics on specificity, delivery vehicle, and kill-switch mechanism. Serif's NHP data readouts and initial indication choices will matter—redosable DNA without integration would be a step change if it actually works. Strand's STX-001 expansion and any new INDs for systemic programmable mRNA are worth tracking. And A2 Bio's regulatory interactions will offer clues about how comfortable the FDA is getting with logic-gated platforms in solid tumors, a notoriously difficult setting for CAR-T.
The "undruggable" gap has been a fixture of oncology for two decades. It probably won't close entirely—biology is stubborn that way. But if DNA-reading therapeutics deliver even a fraction of what their architects promise, the next five years could rewrite which targets are considered tractable, and which cancers are considered treatable.
That's a big if. But it's one that a growing number of biotechs, and the investors backing them, seem willing to bet on.
