Human Tech Tree
Current research2017 · Present (2015 – Oct 2026)

Life / Medicine

Gene Therapy (AAV Vectors)

Harmless virus shells deliver working genes into the body: Luxturna (2017) was the first FDA-approved in vivo gene therapy; ten AAV products are now listed.

Open in the interactive tree →

An adeno-associated virus (AAV) carries a working copy of a gene into a patient's cells without cutting the genome, as CRISPR does. The FDA approved Luxturna (Spark Therapeutics) in December 2017 for an inherited form of blindness caused by RPE65 mutations, the first in vivo gene therapy approved there, and Zolgensma (Novartis) for spinal muscular atrophy in May 2019. Several more AAV products followed.

As of October 2026

The FDA's list of approved cellular and gene therapy products (current as of 17 September 2026) shows ten AAV-based gene therapies, including Luxturna, Zolgensma, Hemgenix, Roctavian, Beqvez, Elevidys and Kebilidi. Safety is the open question: Sarepta reported deaths from acute liver failure in non-ambulatory Elevidys patients in March and June 2025 and added a boxed warning, a third patient died in July 2025 in a trial of a limb-girdle muscular dystrophy therapy, and FDA asked Sarepta to halt US distribution. On 30 July 2025 FDA recommended lifting the voluntary hold for ambulatory patients only, and it revoked the platform designation for Sarepta's AAV technology, which had speeded up reviews of related products.

Open steps

  • Safe high-dose AAV treatment Medium AI leverageAcute liver failure killed two non-ambulatory boys on Elevidys in 2025. A sirolimus cohort of about 25 patients is meant to show whether stronger immunosuppression prevents it.
  • Capsids that reach the brain Medium AI leverageOne IV dose should reach brain neurons and spare the liver. A first such capsid, CAP-002, was cleared for a phase 1/2a trial in 2025; the company's data so far come from primates.
  • How long one dose lasts Medium AI leverageFactor IX stayed stable for four years after a hemophilia B therapy, but the hemophilia A therapy averaged 16 IU/dL factor VIII at one year. What sets durability for each product?
  • Patients with antibodies to the vector Medium AI leverageMany people carry antibodies to AAV. In HOPE-B, 21 of 54 did; their mean factor IX was 34 IU/dL at year 4, but one with a very high titre did not respond. Who can be treated?

Where AI could help

Medium AI leverage. Machine learning already designs better AAV capsids; immune reactions, liver toxicity and manufacturing set the pace.

  • Design capsids that reach one tissue and avoid the liver
  • Predict immune reactions to the vector before patients are dosed
  • Optimise promoters and gene cassettes
  • Screen manufacturing yield before scale-up

Shown so far

  • In August 2024 the Broad Institute's Fit4Function models designed AAV9 capsids, and about 90% of the capsids predicted to work delivered cargo to human liver cells and met five other criteria; models trained on mouse and human cell data predicted behaviour in macaques. source

Prerequisites

Sources

More in Medicine · Present

All 72 points in Medicine →

Open in the interactive tree →