RNA Medicines

Established RNA medicines treat disease by acting on RNA rather than on protein or DNA, and the case for treating them as one platform is that a handful of chemistries — antisense oligonucleotides, small interfering RNA, messenger RNA, self-amplifying and circular RNA, and RNA editing — share a delivery problem, a manufacturing base and a regulatory logic. Established The proof that this is a platform and not a collection of one-offs is the approved-drug ledger: more than a dozen antisense and small-interfering-RNA drugs are licensed, and messenger-RNA vaccines were manufactured and administered in the billions within two years of the pandemic.

Established The single sharpest argument for the platform is a dosing interval. Inclisiran, a small interfering RNA against PCSK9, is given as a subcutaneous injection twice a year to lower LDL cholesterol; a chronic disease managed with a drug taken every six months is a different public-health object from one managed with a daily pill, and that durability — stable chemistry plus a liver-targeting sugar tag that acts as a depot — is what the whole field is selling. Frontier The single sharpest argument against it is a wall: that same liver-targeting chemistry does not reach muscle, lung, tumour or, except by direct injection, the brain, and no modality has cleared that wall at scale.

Established The honest shape of the subject is a proven core and an unproven frontier held apart. Proven: antisense and small interfering RNA work for liver and central-nervous-system targets, and messenger-RNA vaccines work. Frontier Unproven at scale: RNA editing in humans, self-amplifying RNA as a durable therapeutic rather than a vaccine, extrahepatic delivery, and a regulatory pathway for the bespoke n-of-1 antisense drugs the chemistry already makes possible.

Established A note on sourcing. This brief was commissioned in September 2026 from the Institute’s research base. Reading-list entries without links are cited from the bibliographic record rather than re-fetched, and claims are dated no later than early 2026 unless carried by a linked source.

Established The antisense oligonucleotide ledger is the oldest and it defines the modality’s two mechanisms. A gapmer recruits the enzyme RNase H1 to cut a target transcript; a splice-switching oligonucleotide binds pre-messenger RNA to redirect splicing without destroying it. Established Fomivirsen, approved in 1998 for cytomegalovirus retinitis, was the first and was later withdrawn; mipomersen followed in 2013; nusinersen, approved in 2016 for spinal muscular atrophy and dosed by intrathecal injection every four months, is the splice-switching success and the drug that made the modality credible in the central nervous system. Established Eteplirsen, approved the same year for Duchenne muscular dystrophy on the surrogate of restored dystrophin, is the modality’s most contested approval, because the dystrophin increase was small and the clinical benefit disputed. Established Tofersen, approved in 2023 for SOD1-associated amyotrophic lateral sclerosis, extended the pattern and the controversy: it was cleared on a reduction in the biomarker neurofilament rather than on a met clinical endpoint.

Established The small interfering RNA ledger is younger, cleaner and built almost entirely on one delivery trick. Patisiran, approved in 2018 for hereditary transthyretin amyloidosis with polyneuropathy, was the first small interfering RNA drug and used a lipid nanoparticle given intravenously every three weeks. Established Every siRNA drug after it — givosiran for acute hepatic porphyria in 2019, lumasiran for primary hyperoxaluria in 2020, inclisiran for hypercholesterolaemia, and vutrisiran for transthyretin amyloidosis in 2022 — abandoned the lipid nanoparticle for a GalNAc sugar conjugate that carries the drug to liver cells by the asialoglycoprotein receptor. Established The consequence is the durability the platform advertises: vutrisiran is dosed once a quarter and inclisiran twice a year, and in 2025 vutrisiran’s HELIOS-B programme extended it from the nerve to the heart, in transthyretin amyloid cardiomyopathy. Frontier The pattern is unambiguous and it is also the constraint: the modality is a triumph in hepatocytes and largely absent everywhere else.

Established Messenger RNA is the modality that reached everyone, and its record is a vaccine record. The two COVID-19 messenger-RNA vaccines, BNT162b2 and mRNA-1273, were authorised in late 2020 and administered in the billions of doses, the largest deployment of any RNA medicine by orders of magnitude; a messenger-RNA respiratory syncytial virus vaccine followed in 2024. Frontier Beyond prophylaxis, the strongest signal is oncology: a personalised neoantigen messenger-RNA cancer vaccine reported a positive randomised phase 2 result in melanoma when combined with a checkpoint inhibitor. Frontier Protein-replacement messenger RNA — supplying a missing enzyme by repeatedly dosing its transcript — is the harder ambition, because messenger RNA is transient and a chronic deficiency needs chronic redosing, which is the durability problem in reverse.

Frontier Self-amplifying RNA is the first genuinely new modality to reach approval since siRNA, and it is easy to under-rate. A self-amplifying RNA carries an alphavirus replicase that copies the transcript inside the cell, so a much smaller dose produces the same protein output. Established ARCT-154, a self-amplifying COVID-19 vaccine also called zapomeran and marketed as Kostaive, was approved in Japan in late 2023 — the first self-amplifying RNA product anywhere — and later in the European Union. Frontier The counter-signal is commercial rather than scientific: at least one prominent self-amplifying-RNA developer went bankrupt in 2024, a reminder that a working modality and a viable company are different achievements. Speculative Circular RNA, which has no free ends and so resists the exonucleases that degrade linear transcripts, is the next candidate for durable protein expression and remains preclinical, pursued by companies including Orna and Laronde.

Frontier RNA editing is the frontier the field is most excited about and has the least human data on. The approach recruits the cell’s own ADAR enzyme, which converts adenosine to inosine and so rewrites a single RNA base, using a guide oligonucleotide and no exogenous protein; the edit is transient and touches RNA, not the genome. Frontier The first-in-human milestone came from Wave Life Sciences’ WVE-006 for alpha-1 antitrypsin deficiency, which reported the first clinical evidence that recruiting endogenous ADAR could correct the SERPINA1 transcript and restore functional M-type alpha-1 antitrypsin protein. Speculative Whether a transient, redosed RNA edit can restore enough protein durably enough to matter clinically is exactly what a single early programme cannot yet establish, and several other developers — ProQR, Korro Bio, ADARx and the exon-replacement approach of Ascidian — are betting it can.

Established The n-of-1 antisense drug is the modality’s most radical demonstration and its least resolved regulatory object. Milasen was a splice-switching oligonucleotide designed, synthesised and dosed for a single child with a unique CLN7 Batten-disease mutation, from identification to first dose in under a year, and its own authors stated plainly that it is not suited to any other patient. Established It established that a bespoke oligonucleotide can be built for one person; it did not establish how such a drug should be reviewed, funded or supervised, and the traditional randomised trial is undefined for a population of one. Frontier The FDA issued guidance on individualised antisense drugs, and the non-profit n-Lorem Foundation was created to make them for ultra-rare patients, but the pathway remains a set of case-by-case decisions rather than an established category — the same gap the bespoke base editor built for the patient KJ Muldoon exposed on the DNA side.

Established The toxicity ledger differs sharply by modality, which is itself an argument for treating them separately even inside one platform. Antisense oligonucleotides carry class effects — injection-site reactions, thrombocytopenia (severe enough with inotersen to require monitoring), hepatotoxicity, and innate-immune activation through pattern-recognition receptors. Established GalNAc-conjugated siRNA is, by contrast, remarkably well tolerated, which is much of why it has become the default. Frontier Messenger-RNA lipid nanoparticles are reactogenic and carry a rare myocarditis signal seen in the COVID vaccines; and the most serious recent RNA-delivery safety event came from in-body gene editing, where Intellia’s nex-z — a lipid nanoparticle delivering a Cas9 messenger RNA and a guide — caused a grade 4 hepatotoxicity that led the FDA to hold both phase 3 trials on 29 October 2025, with the patient’s death on 5 November 2025 against a denominator of more than 450 of roughly 650 dosed. Established That event is a delivery-toxicity signal, not an editing one, and it lands on every modality that relies on a lipid nanoparticle to the liver.

Frontier The defining open question is extrahepatic delivery, and it is the same question for every modality. GalNAc solved the liver so completely that the field’s centre of gravity moved there; muscle, lung, tumour and the brain remain reachable only by local injection or not at all. Frontier The leading answers are conjugates — antibody-oligonucleotide constructs that ride a tissue receptor, of which the transferrin-receptor approach to muscle pursued by Avidity and Dyne is the most advanced — and next-generation lipid nanoparticles tuned to escape the liver’s default uptake. Speculative None has yet produced an approved extrahepatic systemic RNA drug outside the central nervous system, so the wall is real and current, not historical.

Frontier The second is whether RNA editing graduates from proof-of-concept to durable therapy. The first human evidence that endogenous ADAR can be recruited to restore a functional protein is a genuine milestone, but transient editing implies chronic redosing, and the durability, the editing efficiency in vivo and the off-target A-to-I edits across the transcriptome are all still being measured. Speculative If ADAR editing works durably it is safer than DNA editing precisely because it is reversible; that same reversibility is why it may never be a one-time fix.

Frontier The third is self-amplifying RNA beyond vaccines. A modality that produces more protein per dose is attractive for exactly the protein-replacement and durable-expression problems messenger RNA struggles with, but self-amplification also prolongs innate-immune activation, and the therapeutic (as opposed to vaccine) window is unproven. Speculative Circular RNA is the parallel bet on durability by a different route, and it has no clinical data at all.

Frontier The fourth is the regulatory shape of the n-of-1 drug. The chemistry to build a bespoke antisense drug for one patient exists and has been used; what does not exist is a pathway that says how safety, efficacy and payment are established when the trial population is one. Handwave Every proposal to scale n-of-1 medicine presupposes that pathway, and none of them has it.

Established The binding bottleneck is delivery outside the liver, and it decomposes into two physical problems. The first is tissue targeting: getting the oligonucleotide or nanoparticle to the intended cell type rather than to the hepatocytes that take up almost everything by default. Established The second is endosomal escape: even a molecule that reaches the right cell is trapped in an endosome, and the widely cited figure is that under two per cent of internalised siRNA escapes into the cytoplasm where it can act. Frontier Those two together are why the liver, which is fenestrated and richly endocytic, was solved first and why everything else is hard; the GalNAc-siRNA success is as much an accident of hepatic biology as a triumph of chemistry.

Frontier The second bottleneck is durability for the messenger-RNA modalities. A single messenger-RNA dose produces protein for days; a chronic disease needs the protein for years, so protein-replacement messenger RNA implies indefinite redosing with all the reactogenicity that entails. Frontier Self-amplifying and circular RNA are the two engineering answers, trading, respectively, prolonged innate activation and manufacturing complexity for longer expression.

Established The third is innate immunogenicity, which cuts both ways. The same RNA that a vaccine wants to be seen by the immune system is a liability for a therapeutic that must be given repeatedly, and the chemical modifications that quiet innate sensing — the modified nucleosides that made therapeutic messenger RNA possible at all — are the modality’s enabling technology and its patent battleground. Frontier Getting the immune-visibility dial right for a chronic therapeutic is unsolved for the amplifying modalities.

Established This subject consumes delivery chemistry that a neighbouring field builds for its own reasons. The lipid-nanoparticle architecture that carries messenger RNA — the four-component design and the ionisable-lipid lineage from MC3 to SM-102 and ALC-0315 — is the object of study in Nanomedicine, and every extrahepatic advance this brief waits on will come from that chemistry rather than from the RNA itself. Established The GalNAc conjugate and the emerging antibody-oligonucleotide conjugates are the same story: the payload is solved and the vehicle is the constraint.

Established Two dependencies are shared with genome engineering and visible in its record. In-body genome editing is, mechanically, an RNA-delivery problem — a lipid nanoparticle carrying a Cas9 messenger RNA and a guide, or a base-editor transcript — so the delivery, manufacturing and hepatotoxicity lessons assessed in Genetic Engineering are this subject’s lessons too. Frontier The n-of-1 regulatory precedent runs on both sides of the RNA-DNA line at once, which is why the bespoke antisense drug and the bespoke base editor face the same institutional gap.

Speculative The dependency nobody funds is a predictive model of extrahepatic uptake. The field advances tissue by tissue empirically, and there is no validated way to predict from a conjugate’s structure how much of it will reach, say, cardiac muscle at a tolerable dose. Handwave Until that model exists, extrahepatic delivery remains a screening exercise rather than a design one.

Frontier The decisive test is a single demonstration of durable, well-tolerated RNA delivery to a therapeutic target outside the liver and central nervous system — muscle is the nearest — at a dose that restores or suppresses the target protein for months from one administration. Everything the platform promises beyond hepatology and intrathecal neurology rests on clearing that wall, and no approved systemic drug has. Frontier The antibody-oligonucleotide conjugates aimed at muscle are the experiment in progress, and a clean, durable extrahepatic result would expand the addressable disease space more than any new chemistry.

Frontier The second experiment is durable protein restoration by RNA editing in humans. The first-in-human ADAR result showed the mechanism works; what would settle the modality is a demonstration that redosed editing holds protein at a therapeutic level over a year with an acceptable off-target profile. Speculative That is a measurement the current programmes are set up to make and have not yet reported.

Frontier The third is institutional: an n-of-1 genetic-medicine approval pathway that has been used more than once. Milasen proved a bespoke oligonucleotide can be built; a pathway that reviews, funds and supervises a population-of-one drug as a category, rather than as a series of compassionate-use exceptions, is the policy result that would convert n-of-1 medicine from heroic to routine. Handwave No agency has established one, and until one exists the model does not scale.

Established The central engineering systems are the vehicle and the manufacturing line, not the RNA. The oligonucleotide chemistry — phosphorothioate backbones, 2'-modified sugars, the GalNAc conjugate — is mature; the engineering frontier is the delivery vehicle for everything the GalNAc route cannot reach, which means antibody-oligonucleotide conjugates and tissue-tropic lipid nanoparticles. Frontier Each new target tissue is, in effect, a new vehicle-engineering programme, which is why the platform generalises slowly despite the payload being trivial to reprogram.

Established Manufacturing is the modality’s quiet advantage. Messenger RNA is made by in-vitro transcription from a DNA template, a cell-free process that can be reprogrammed to a new sequence in days and scaled fast — the property the COVID response demonstrated at civilizational scale, when billions of doses of a novel sequence were produced within a year. Frontier Oligonucleotide manufacturing by solid-phase synthesis is more constrained and its raw-material supply is a real bottleneck, but nothing in RNA manufacturing resembles the one-patient-at-a-time constraint that limits cell therapy.

Speculative The reprogrammability is the platform’s deepest engineering claim: the same line makes any sequence. If a delivery vehicle for a given tissue is solved once, every RNA drug for that tissue inherits it, so the field’s progress is lumpy — nothing, then a whole tissue at once. Handwave That is the optimistic reading of the extrahepatic wall: it is a small number of hard vehicle problems standing in front of a large number of easy payload problems.

Established The nearest neighbour is delivery, and it sets this subject’s ceiling. Every extrahepatic advance depends on the lipid-nanoparticle and conjugate chemistry assessed in Nanomedicine, and the messenger-RNA lipid nanoparticle is the single most consequential object shared between the two subjects. Frontier The traffic runs both ways: RNA payloads are what made the lipid nanoparticle a clinical technology rather than a laboratory curiosity.

Frontier Genome engineering is the second adjacency, and the boundary between them is blurring. In-body CRISPR and base editing are delivered as RNA in a lipid nanoparticle, so the in-vivo editing programmes assessed in Genetic Engineering are, at the delivery layer, RNA medicines; the difference is whether the RNA acts on a transcript or installs a permanent change. Frontier RNA editing sits deliberately between them: the durability of a drug with the reversibility of a transcript.

Frontier Two further adjacencies are load-bearing. The central-nervous-system oligonucleotides — nusinersen, tofersen — are the working half of the delivery story and are assessed as therapies in Neurogenetics, where intrathecal injection substitutes for the systemic delivery this brief says is unsolved. Frontier And the messenger-RNA and self-amplifying-RNA vaccine platforms are the subject of Universal Vaccines, which owns the immunological questions this brief treats only as a delivery-and-durability problem.

Established The first institutional fact is that RNA medicines fit the existing drug system better than cell or gene therapies do. They are manufactured, shipped, stored and dosed like conventional biologics, they are redosed rather than given once, and they are therefore paid for from recurring budgets that health systems already understand. Frontier Inclisiran’s twice-yearly interval was pitched partly as an adherence and population-health tool — a clinic-administered injection removes the daily-pill compliance problem — which is a health-system argument as much as a pharmacological one, and its real-world cost-effectiveness at population scale is still being established.

Established The second is the missing n-of-1 pathway. Milasen and the individualised-antisense guidance that followed created precedent but not a category; a bespoke drug for one patient has no trial, no obvious payer and no standing supervisory framework. Frontier The n-Lorem Foundation exists precisely because the market cannot serve a population of one, and philanthropic manufacture is a workaround, not an institution.

Established The third is manufacturing sovereignty, which the pandemic turned from abstraction into policy. The demonstration that a novel messenger-RNA sequence could be designed and manufactured in billions of doses within a year made RNA-manufacturing capacity a strategic asset, and several states now treat it as one. Speculative That capacity is the closest thing the platform has to a moat, and it is institutional rather than scientific.

Frontier The characteristic ethical problem of RNA medicine is the n-of-1 drug: real benefit for one identifiable child against an evidentiary and distributive framework built for populations. A bespoke oligonucleotide can help a single patient with a unique mutation, but the resources, expertise and manufacturing to build it are available to almost no one, and choosing which ultra-rare patient gets a bespoke drug is a rationing decision with no established process. Speculative A medicine that can, in principle, be made for anyone with a defined mutation, but in practice for almost no one, is a new kind of inequity, not a familiar one.

Established The durability advantage carries an ethical edge. A drug given twice a year through a clinic can improve outcomes for populations that struggle with daily adherence, which is a genuine equity benefit; it also concentrates the decision and the cost in the health system rather than the patient, and its price determines whether that benefit is realised or merely advertised. Frontier The gene-therapy price wall — one-time products listing between roughly two and four million dollars — is the comparator RNA medicines are measured against, and their redosed, lower-per-dose economics are part of the argument for them.

Established The safety asymmetry between prevention and therapy is an ethical fact, not just a clinical one. A reactogenic profile acceptable in a one-or-two-dose vaccine is a different proposition in a drug taken for life, and the rare myocarditis signal in the COVID messenger-RNA vaccines is the reference case for how a small absolute risk is weighed against a large benefit. Frontier Transposing vaccine-era risk tolerance onto chronic therapeutics would be a category error the field must avoid.

Speculative The civilizational claim is that RNA becomes a general programmable-medicine platform: any protein raised or lowered, any transcript edited, in any tissue, from one manufacturing base. The payload is already trivial to reprogram and the manufacturing is already fast; the only thing standing between the current liver-and-vaccine reality and that general platform is extrahepatic delivery. Frontier That is an unusually clean statement of a field’s bottleneck, and it is why the delivery problem, not the RNA, is the whole game.

Established The pandemic already ran the civilizational-scale experiment on manufacturing. A previously niche modality was scaled to billions of doses of a novel sequence within a year, which is the strongest evidence any RNA claim has that the manufacturing side of a programmable-medicine future is achievable rather than aspirational. Frontier No comparable demonstration exists for the delivery side, which is the asymmetry the whole subject turns on.

Frontier Declare the balance honestly. For: more than a dozen approved drugs, a twice-yearly cholesterol therapy, a self-amplifying vaccine on the market, the first human RNA edit, and a manufacturing base proven at planetary scale. Established Against: the entire systemic modality is confined to the liver plus intrathecal neurology, messenger-RNA therapeutics are transient, RNA editing has one early programme, and the n-of-1 model has no pathway. Frontier The honest verdict is that RNA medicine is a proven, durable, well-manufactured platform whose reach is bounded by a single unsolved problem, and that the next decade is a referendum on delivery.

These horizons track results that would change what can honestly be said about RNA as a therapeutic platform, not predictions of clinical availability.

• 10 yr: Frontier The extrahepatic wall is the link most likely to move: antibody-oligonucleotide conjugates to muscle report their first approvals or fail, and either outcome resets the addressable disease space. Frontier RNA editing produces its first durable human protein-restoration data, and self-amplifying RNA is tested as a therapeutic rather than only a vaccine; expect the n-of-1 pathway to remain a set of exceptions unless a regulator makes it a category.
• 25 yr: Speculative Either general extrahepatic delivery is solved for two or three major tissues, in which case RNA becomes the default modality for a large class of chronic diseases, or it is not, and the platform consolidates as hepatology, intrathecal neurology and vaccines. Speculative Circular and self-amplifying RNA either make durable protein expression routine or remain niche around the transient-messenger-RNA problem.
• 50 yr: Speculative If delivery and durability are both solved, programmable RNA medicine merges with in-body editing into a single reprogrammable-therapeutics discipline in which raising, lowering or rewriting any transcript in any tissue is a routine act. Handwave The alternative endpoint is equally coherent: RNA remains a superb tool for a defined set of tissues and targets, and the general-platform language is quietly retired.
• 100 / 250+ yr: Handwave At this horizon the interesting question is whether the transcriptome becomes a continuously tunable control surface — read, adjusted and re-read as a matter of routine maintenance. Handwave Every component of that is asserted rather than demonstrated, and it is worth ending on the observation that the platform’s hardest present constraint, getting a molecule into a cell that is not a hepatocyte, has been true for the entire history of the field.

• Depends on This subject waits on results other briefs are producing for their own reasons. It consumes lipid-nanoparticle and conjugate delivery chemistry from Nanomedicine, in-body editing and manufacturing lessons from Genetic Engineering, and the intrathecal-delivery and CNS-target work assessed in Neurogenetics. None is a result this subject can produce for itself.
• Requires (not on this map) Five constraints bind harder than any single chemistry. Reliable delivery outside the liver and CNS is the scientific result the whole platform waits on. Population-scale messenger-RNA manufacturing exists, proven by the pandemic, and is the platform’s strongest asset. A regulatory pathway that treats a population-of-one drug as a category, not an exception, is absent. GMP lipid and GalNAc-conjugate supply is a genuine raw-material constraint on oligonucleotide scale-up. And a reimbursement model for a clinic-administered, twice-yearly chronic therapy determines whether the durability advantage reaches patients.
• Enables What this subject supplies is a reprogrammable therapeutic layer: any protein raised, lowered or rewritten once a delivery vehicle for its tissue exists, and a manufacturing base that can produce a new sequence in days. No typed enabling edge is recorded, because that contribution is a capability others build on rather than a discrete technology.
• Adjacent The adjacent subjects share the vehicle and the boundary. Nanomedicine owns the delivery chemistry that decides the platform’s reach. Genetic Engineering shares the in-body RNA-delivery layer and the n-of-1 precedent. Neurogenetics is where the CNS oligonucleotides are assessed as therapies, and Universal Vaccines owns the immunology of the messenger-RNA and self-amplifying-RNA platforms.

Established “Messenger-RNA vaccines change your DNA.” Messenger RNA acts in the cytoplasm, is not reverse-transcribed by human cells, and is degraded within days; it never enters the nucleus and does not alter the genome. Established The transience that makes it safe on this point is the same transience that makes protein-replacement messenger RNA hard, which is the honest trade the modality carries.

Frontier “RNA editing is just reversible CRISPR.” ADAR editing rewrites a single RNA base, adenosine to inosine, using the cell’s own enzyme and no permanent change to DNA; it cannot make arbitrary edits, it is transient, and it must be redosed. Speculative That is a genuinely different object from genome editing, safer because reversible and weaker for the same reason, and conflating the two flatters both.

Established “siRNA can silence a gene in any tissue.” Approved systemic siRNA drugs work in the liver because the GalNAc conjugate targets hepatocytes; outside the liver, delivery is the unsolved problem, and under two per cent of internalised siRNA escapes the endosome even when it reaches the right cell. Frontier The modality’s reach is a fact about delivery chemistry, not about RNA interference, which works everywhere in a dish and almost nowhere outside the liver in a patient.

Frontier “Self-amplifying RNA is still experimental.” A self-amplifying RNA vaccine has been approved and marketed since late 2023, so the modality is not hypothetical; what is unproven is its use as a durable therapeutic rather than a vaccine. Speculative The distinction between an approved vaccine modality and an unproven therapeutic one is exactly the distinction this brief is built to preserve.

Frontier “n-of-1 antisense drugs are a scalable cure for rare disease.” Milasen showed a bespoke oligonucleotide can be built for one patient; it did not show how such a drug is reviewed, funded or supervised, and no pathway treats a population of one as a category. Handwave Calling n-of-1 medicine scalable is the step where the argument works by assertion, because the chemistry scales and the institution does not.

Established “RNA medicines are cheap because they are simple molecules.” The molecule is simple and the delivery, the durability engineering and the ultra-rare economics are not; inclisiran’s value case rests on adherence and population health, not on a low unit price, and the n-of-1 drugs have no conventional price at all. Frontier The RNA is the cheap part, and it is not the part that determines what these drugs cost or who receives them.