CRISPR Gene Editing and AI Diagnosis for Rare Genetic Diseases: 2026 Guide

The first CRISPR medicine, Casgevy, is now approved for sickle cell disease and beta thalassaemia, and AI tools are helping families get a genetic diagnosis faster. Here is what is real, what is still experimental and what it involves.

The problem today: long waits for answers, few treatments that fix the cause

A disease is called rare when it affects only a small number of people. Each one is uncommon, but together they are not. The European Parliament's research service estimates that there are more than 6,000 rare diseases, affecting about 300 million people worldwide and about 30 million in the European Union. Around 72% of them have a genetic origin, which means they are caused by a change (a "variant" or "mutation") in a person's DNA, the instruction manual inside almost every cell.

Two problems have shaped life with a rare genetic disease for decades.

• The diagnostic odyssey. This is the name families use for the long search for a diagnosis. The same European Parliament briefing says patients typically wait about five years. During that time, people may see many specialists, receive wrong diagnoses and have tests that do not give an answer.
• Treatments that manage symptoms, not the cause. Even with a diagnosis, most rare genetic diseases have no approved medicine that corrects the faulty instruction. Care often means regular blood transfusions, pain control, special diets or, for some, a bone marrow or liver transplant.

Two blood disorders show both the burden and the change that is now happening. In sickle cell disease, a variant in the gene for haemoglobin (the protein in red blood cells that carries oxygen) makes red cells stiff and curved. These cells block small blood vessels and cause sudden, severe episodes of pain called vaso-occlusive crises, along with long-term damage to organs. In transfusion-dependent beta thalassaemia, the body makes too little working haemoglobin, so people need blood transfusions every few weeks for life, and iron from those transfusions builds up in the heart and liver.

This guide covers both halves of the story: gene-editing treatments that change DNA, starting with the first approved one, and artificial intelligence (AI) tools that help doctors find the genetic cause of a rare disease faster.

How gene editing and AI diagnosis work

CRISPR-Cas9: cutting DNA at a chosen spot

CRISPR-Cas9 is a tool borrowed from bacteria. It has two parts. A short "guide" molecule finds a matching stretch of DNA, and an enzyme called Cas9 cuts both strands of the DNA at that spot. The cell then repairs the cut, but its repair is often imperfect and leaves small changes. Scientists use this to switch off a specific piece of DNA.

Casgevy (exagamglogene autotemcel, often shortened to exa-cel) uses this approach in a clever way. Before birth, everyone makes fetal haemoglobin, a form that works well and does not sickle. After birth, a gene called BCL11A switches fetal haemoglobin off. Casgevy cuts a small control region that turns BCL11A on in red blood cell precursors. With that switch disabled, the cells make fetal haemoglobin again, which protects against sickling and makes up for the missing haemoglobin in thalassaemia.

Base editing: changing a single DNA letter

DNA is written in four chemical "letters" (A, C, G and T). Base editing changes one letter into another without cutting through both strands of DNA. It uses a disabled form of Cas9 that only nicks one strand, joined to an enzyme that chemically rewrites a letter, for example an A into a G. Because there is no full cut, researchers hope it causes fewer unwanted changes. Beam Therapeutics' experimental BEAM-101 (risto-cel) uses base editing to copy natural, harmless DNA changes that keep fetal haemoglobin switched on in some healthy people.

Prime editing: rewriting a short passage

Prime editing is a newer method that can make a wider range of small corrections, including inserting or removing a few letters. It carries its own template of the correct sequence and writes it into the DNA. In 2025, Prime Medicine reported the first human data for prime editing, in a patient with chronic granulomatous disease, an inherited immune disorder.

Two ways to deliver the editor

• Outside the body (ex vivo). Doctors collect the patient's own blood stem cells (the cells in bone marrow that make all blood cells), edit them in a specialised laboratory and give them back through a drip. Casgevy works this way.
• Inside the body (in vivo). The editing instructions are packed into tiny fat bubbles called lipid nanoparticles and given as an infusion. These particles mainly travel to the liver, so this route currently suits diseases driven by the liver. No cells are removed and no chemotherapy is needed.

AI for diagnosis: finding the one variant that matters

Genome sequencing reads all or most of a person's DNA. Each person carries thousands of variants that differ from a reference genome, and almost all are harmless. The hard part is finding the one that explains the illness. AI helps in three main ways:

• Scoring variants. Programs such as Google DeepMind's AlphaMissense predict whether a single-letter change in a gene is likely to be harmful.
• Ranking candidates. Tools such as AI-MARRVEL combine a patient's sequencing data with a standard list of their symptoms (called Human Phenotype Ontology terms) and rank the variants most likely to be the cause.
• Reading facial features. Some genetic syndromes cause subtle, typical facial features. Tools such as GestaltMatcher compare a patient's photo with photos of people who have confirmed diagnoses.

The main treatments and tools, and where they stand

Casgevy (Vertex Pharmaceuticals and CRISPR Therapeutics)

Casgevy is the first approved medicine made with CRISPR. It is a one-time treatment that edits a patient's own blood stem cells. It is for people with severe sickle cell disease who have repeated pain crises, and for people with transfusion-dependent beta thalassaemia. The UK medicines regulator (MHRA) approved it first, in November 2023, for people aged 12 and over. The US Food and Drug Administration (FDA) approved it for sickle cell disease in December 2023 and for thalassaemia in January 2024, and in July 2026 widened its use to children aged 2 and over. According to the Thalassaemia International Federation, it is approved in 39 countries.

Lonvoguran ziclumeran, or lonvo-z (Intellia Therapeutics)

Lonvo-z (also called NTLA-2002) is an in vivo CRISPR treatment for hereditary angioedema, a rare inherited condition that causes sudden, painful and sometimes dangerous swelling. It is a single infusion that switches off a liver gene (KLKB1) involved in the swelling pathway. It is not yet approved anywhere. In September 2026 the FDA accepted Intellia's application for priority review, with a decision target date of 10 March 2027.

Risto-cel, or BEAM-101 (Beam Therapeutics)

Risto-cel is an experimental base-edited cell therapy for severe sickle cell disease. Like Casgevy, it uses the patient's own stem cells and needs chemotherapy before the cells are returned. Beam has finished dosing adults and adolescents in its BEACON study and says it may apply for US approval as early as the end of 2026.

Personalised ("N-of-1") editing

N-of-1 means a treatment made for one specific patient. In 2025, doctors at Children's Hospital of Philadelphia and Penn Medicine treated a baby known as KJ, who had a severe form of CPS1 deficiency, a liver disorder in which toxic ammonia builds up in the blood. The team designed a base editor for his exact variant, made it in about six months and delivered it in lipid nanoparticles. In the UK, Great Ormond Street Hospital and University College London used base-edited immune cells (CAR7 T cells) to treat children with a hard-to-treat leukaemia, starting with a teenager named Alyssa in 2022.

AI diagnostic tools

• AlphaMissense (Google DeepMind, 2023): a public catalogue of predicted effects for about 71 million possible single-letter protein changes. It is a research resource, not a clinical test.
• AI-MARRVEL (Baylor College of Medicine and Texas Children's Hospital, 2024): ranks candidate variants for Mendelian disorders, which are conditions caused by a change in a single gene.
• GestaltMatcher (University of Bonn and partners, 2022): facial analysis linked to the Face2Gene platform used by clinical geneticists.

What the evidence shows for gene therapy

Casgevy in sickle cell disease (CLIMB SCD-121)

The main study was published in the New England Journal of Medicine in 2024. It included people aged 12 to 35 with severe sickle cell disease. Of 30 people who had been followed long enough, 29 had no severe pain crises for at least 12 months in a row, and all 30 avoided hospital stays for crises over that period. Fetal haemoglobin rose to over 40% of total haemoglobin. Limitations: the study had no comparison group, it was small, and follow-up is still only a few years.

Casgevy in beta thalassaemia (CLIMB THAL-111)

In the companion study, 32 of 35 evaluable patients stopped needing transfusions for at least 12 months in a row while keeping a healthy haemoglobin level. Responses were seen across different genetic types, including the most severe forms.

Casgevy in children

The data behind the US expansion came from children aged 5 to under 12. In sickle cell disease, all 8 children who could be assessed had no severe pain crises for at least 12 months. In thalassaemia, 8 of 9 assessable children stopped transfusions for at least 12 months. There were no trial results for children aged 2 to 4; the FDA based that part of the decision on the product's characteristics and the available paediatric data.

Lonvo-z in hereditary angioedema (Phase 2 and HAELO Phase 3)

In the Phase 2 study published in the New England Journal of Medicine, a single infusion lowered attack rates compared with placebo (a dummy infusion). In the Phase 3 HAELO trial of 80 patients, people given a 50 mg dose had 87% fewer monthly attacks than the placebo group, and 62% were free of both attacks and other preventive treatment over six months, compared with 11% on placebo. Side effects reported were mild or moderate. Limitations: long-term data for permanent liver editing in larger groups are still limited.

Base editing: BEACON and BE-CAR7

Early BEACON results for risto-cel showed high levels of fetal haemoglobin. One participant died after treatment; Beam reported that the death was linked to the busulfan chemotherapy used to prepare the bone marrow, not to the edited cells. In the UK BE-CAR7 leukaemia study, published in December 2025, 9 of 11 patients (82%) reached a very deep remission that allowed a stem cell transplant, and 64% remain free of disease.

Personalised editing: baby KJ

KJ received three doses between February and April 2025. His doctors reported no serious side effects, better tolerance of protein in his diet and less need for medicine to remove ammonia. This is one child, and the team itself says longer follow-up is needed.

AI diagnosis

• AI-MARRVEL was tested on three independent groups totalling 1,377 patients. It outperformed four established tools (Exomiser, LIRICAL, PhenIX and Xrare) and roughly doubled the number of solved cases compared with them. It worked less well for recessive conditions (where a person needs two faulty copies of a gene) and was trained mainly on variants inside genes, so its performance on other DNA regions is unclear.
• AlphaMissense classified 89% of about 71 million possible variants as likely harmful or likely harmless, with about 90% precision on a database of known disease variants. Its developers say it is not designed to be used directly in the clinic.
• GestaltMatcher was built from 17,560 patient photos covering 1,115 rare diseases. It can suggest a match even when only a couple of earlier patients with that condition are known. Accuracy depends on the photo and the population in the reference database.

Who can benefit from gene therapy

Gene editing is not a general option for all rare diseases. Today it is only possible when the exact genetic cause is known and there is a way to deliver the editor to the right cells.

Casgevy may be considered for:

• People with severe sickle cell disease and repeated pain crises, or with transfusion-dependent beta thalassaemia.
• Those within the approved age range in their country (2 and over in the US; 12 and over in the EU; 12 and over in the UK, with a reported extension to 5 and over).
• People well enough to tolerate strong chemotherapy and a stay in a stem cell transplant unit.
• In England, NICE managed access for sickle cell disease covers people aged 12 and over with at least two crises a year in the previous two years, who would be suitable for a stem cell transplant but have no matched related donor.

It is usually not suitable for:

• People with serious damage to the liver, heart, lungs or kidneys that makes chemotherapy too risky.
• People who have had certain earlier gene therapies, or who have some active infections. Exact rules differ between centres.
• People with milder disease that is well controlled with standard care, where the risks may outweigh the benefits.

AI-supported diagnosis is most useful for children and adults with unexplained symptoms that suggest a genetic cause, such as developmental delay, several organs affected at once or a family history, especially when earlier tests were negative. These tools are used by genetics teams, not bought directly by patients.

What to expect as a patient

For Casgevy, the whole process usually takes many months from first referral to recovery at home. Typical steps:

1. Referral. Your haematologist (blood specialist) refers you to an authorised treatment centre, a hospital approved to give the therapy.
2. Assessment. Tests check your heart, lungs, liver, kidneys and general health. You discuss fertility, because the chemotherapy can cause infertility, and you may be offered egg, sperm or tissue storage first.
3. Preparing the blood. People with sickle cell disease often have blood transfusions for a period beforehand to lower the share of sickled cells.
4. Stem cell collection. Medicines move stem cells from the bone marrow into the blood. A machine then collects them over several hours (a process called apheresis). Some people need more than one collection cycle.
5. Manufacturing. The cells are sent to a specialised laboratory, edited, tested and frozen. This takes weeks to months.
6. Chemotherapy. You are admitted to hospital and receive busulfan, a strong chemotherapy that clears space in the bone marrow for the edited cells.
7. Infusion. The edited cells are given back through a drip, much like a blood transfusion.
8. Recovery in hospital. You stay in hospital, often for several weeks, until the new cells start making blood. During this time your immunity is very low and you may need transfusions and antibiotics.
9. Follow-up. Regular clinic visits continue for months, and regulators require long-term monitoring for many years.

For an in vivo treatment like lonvo-z, the experience in trials is very different: a single outpatient infusion, with no stem cell collection and no chemotherapy.

For an AI-supported genetic diagnosis, the patient experience is simple: a blood or saliva sample for sequencing, a detailed record of symptoms and sometimes a facial photo. The AI analysis happens in the laboratory. Results still need to be reviewed by experts, and confirmation tests may be needed before a final diagnosis.

Gene therapy: access and cost in the US, UK and EU

United States

The US list price of Casgevy is $2.2 million per treatment. Coverage comes through private insurers and Medicaid, the public programme for people on low incomes. The Centers for Medicare & Medicaid Services runs a Cell and Gene Therapy Access Model for sickle cell disease in Medicaid, and the Thalassaemia International Federation reports reimbursement arrangements in the US. Your treatment centre's financial team can check your specific insurance.

United Kingdom

NICE, the body that decides which treatments the NHS in England funds, recommended Casgevy for transfusion-dependent beta thalassaemia in 2024 (guidance TA1003) and for severe sickle cell disease in 2025 (guidance TA1044). Both recommendations use managed access, which means the NHS funds the treatment while collecting more data. The Thalassaemia International Federation reports agreements covering Wales, Scotland and Northern Ireland as well. The NHS price is confidential; there is no official data on it.

European Union

The European Commission granted a conditional marketing authorisation in February 2024, which means the medicine can be sold across the EU while further data are collected. Funding is decided country by country. Reported milestones include early access programmes in France (2024) and Italy (2024), reimbursement agreements in Italy and Denmark (2025) and Austria (reported 2025), and a German agreement with the statutory health insurance association (GKV-Spitzenverband) in June 2026. There is no official data on prices in the EU.

Elsewhere and for international patients

Casgevy is also approved in Canada, Switzerland, Saudi Arabia, Bahrain, the United Arab Emirates and Qatar, among 39 countries in total. Treatment is only given at authorised centres; the network grew to more than 75 centres worldwide by August 2025.

If you are thinking of travelling for treatment, check carefully:

• Approval in that country. Make sure the treatment is approved there for your age and condition, or that you would join a registered clinical trial.
• Centre experience. Ask whether the hospital is an authorised treatment centre and how many patients it has treated.
• Records to bring. Genetic test reports, transfusion and crisis history, organ test results, current medicines and vaccination records, translated if needed.
• Time abroad. Plan for several months, including stem cell collection, manufacturing time, hospital stay and early follow-up.
• Follow-up at home. Agree in writing who will monitor you after you return, because long-term checks are required.
• Insurance and funding. Get written confirmation of what is covered, including complications and repeat collections.

Risks, limits and realistic expectations

• Chemotherapy is the biggest risk for ex vivo treatments. Busulfan causes very low blood counts, mouth sores, infections, hair loss and possible infertility. It can also cause serious liver and lung problems. The death in the BEACON study was attributed to this chemotherapy.
• Unintended edits. Editors can sometimes change DNA in the wrong place or cause larger DNA rearrangements. In the Casgevy trials, no off-target edits were found, but very long-term effects cannot yet be known.
• Delivery is a barrier. Lipid nanoparticles mainly reach the liver, so many diseases of the brain, muscles or heart cannot yet be treated this way. Viral delivery has risks: in 2022, a young man with Duchenne muscular dystrophy died after a high-dose, experimental gene-editing treatment delivered by a virus, and a later study linked the death to a strong immune reaction to the virus.
• Existing damage may not reverse. Gene editing can stop new crises or transfusions but may not undo earlier damage to bones, kidneys, lungs or the brain.
• Long follow-up. Because edits are permanent, regulators require years of monitoring.
• AI limits. AI suggestions can be wrong, may perform less well for under-represented populations and still need expert confirmation. A "likely harmful" score alone is not a diagnosis.

Clinical trials and what is next for rare genetic diseases

• Casgevy follow-up: a long-term study (CLIMB-131) is following treated people of all ages to check safety and durability.
• Risto-cel (BEAM-101): the BEACON trial (NCT05456880) has finished dosing adults and adolescents; Beam expects updated data and a possible US filing around the end of 2026.
• BEAM-302: an in vivo base editor for alpha-1 antitrypsin deficiency, an inherited condition that damages the lungs and liver. The first patient in a pivotal group was dosed in July 2026.
• Lonvo-z: FDA decision expected by 10 March 2027.
• Prime editing: PM359 gave the first human proof that prime editing can work, but Prime Medicine stopped developing it in 2025 for business reasons, so it is not currently moving forward.
• Personalised editing: the KJ case has encouraged work on faster, reusable "platform" approaches for very rare liver disorders, but these remain experimental.

To search for trials, use ClinicalTrials.gov, the EU Clinical Trials Information System or the UK's Be Part of Research website, with terms such as "CRISPR", "base editing" and your condition. Patient organisations such as EURORDIS and the Thalassaemia International Federation also share trial news.

Questions to ask your doctor about gene therapy

These questions can help you prepare for an appointment:

1. Has my genetic cause been confirmed, and would genome sequencing or reanalysis of earlier results help?
2. Are AI-supported tools used in your genetics laboratory, and could my old data be looked at again?
3. Am I eligible for Casgevy or a gene-editing trial, given my age, disease severity and organ health?
4. How does gene editing compare with a stem cell transplant from a donor, if I have a matched sibling?
5. What are my personal risks from the chemotherapy, including fertility, and what fertility preservation is available?
6. Which authorised treatment centre would I go to, and how many patients has it treated?
7. How long will the whole process take, and how long will I be in hospital?
8. Who pays, and what would I need to cover myself?
9. What follow-up will I need, and for how many years?
10. What results can I realistically expect, and what damage might not improve?

Questions patients ask