CAR-T cell therapy
A patient's own immune cells are re-engineered to hunt cancer and can cure some blood cancers.
Open in the interactive tree →T cells are taken from the blood, given a gene for a chimeric antigen receptor that recognizes a tumor marker, multiplied and re-infused. The FDA approved the first CAR-T drugs in 2017 (Kymriah on 30 August, Yescarta on 18 October); by the end of 2024 seven products were approved, all for blood cancers. Some patients with otherwise fatal leukaemia and lymphoma remain in remission for years.
As of October 2026
Researchers are extending CAR-T to autoimmune diseases: after first lupus results from Erlangen in 2022 (five patients), trials now include systemic sclerosis, myositis and rheumatoid arthritis, although data remain limited. A major focus is 'in vivo' CAR-T, which delivers the CAR gene directly into a patient's T cells so that no cells have to be harvested and shipped (reportedly about 20 percent of US patients die before their custom product is ready); first clinical reports are appearing, for example a lentiviral CD19 approach in neurological autoimmune disorders (NEJM, September 2026). Solid tumors remain largely out of reach, with Tecelra, a T-cell-receptor therapy for synovial sarcoma (August 2024), the first engineered cell therapy approved for a solid tumor.
Open steps
- In-body CAR-T without cell factories Medium AI leverageDelivering the CAR gene inside the body skips harvesting and shipping, but T-cell-only targeting, dose control and long-term safety remain unproven.
- Safe targets in solid tumours High AI leverageFew antigens appear on tumour cells but not on vital tissues; AI must find targets and design binders that spare healthy organs.
- Cells that stay active in tumours Medium AI leverageIn solid tumours CAR-T cells tire and are suppressed; armoured designs, metabolic edits and combination plans must keep them working.
- Predicting severe side effects Medium AI leverageCytokine storm and neurotoxicity are dangerous and hard to predict; early warning from labs and monitoring could allow outpatient use and cut costs.
Where AI could help
Medium AI leverage. AI can design binding domains and predict toxicity, but solid tumors, manufacturing and safety need lab and clinical work.
- Designing binders for new tumor targets and escape variants
- Predicting toxicity risk such as cytokine release
- Designing delivery vectors for in vivo CAR gene transfer
- Quality control of cell manufacturing
Shown so far
- A Bits to Binders competition tested about 12,000 AI-designed binders in functional CAR-T screens, with team success rates from 0.6% to 38.4% (webinar announced for June 2026). source
- In November 2024 a Munich preprint reported a CAR-T cell with an RFdiffusion-designed BCMA binder that was active in a mouse model of multiple myeloma (preprint, not peer reviewed). source
Prerequisites
- Chemotherapy1942
- Cell culture and HeLa cells1951A patient's T cells are grown and engineered in culture before being returned
- Genetic engineering1973