Universal flu vaccine
Flu shots must be remade every year and still miss; a vaccine that covers all flu strains for years does not exist.
Open in the interactive tree →Influenza viruses change their surface proteins so fast that vaccines are updated twice a year and sometimes miss the circulating strain. A universal vaccine would target parts of the virus that barely change, or train both antibodies and T cells to hit many strains. Several designs work in animals, but a 2019 efficacy trial of one T-cell candidate in 2,152 adults found no benefit.
As of October 2026
WHO estimates about one billion flu cases a year, 3 to 5 million severe cases and 290,000 to 650,000 respiratory deaths. In May 2025 HHS and NIH launched Generation Gold Standard, reportedly US$500 million for whole-virus inactivated candidates with human trials planned in 2026 and a US approval target of 2029; critics note earlier universal-vaccine attempts failed. Centivax began a phase 1a trial of its pan-influenza candidate Centi-Flu 01 in Australia in February 2026 and said in May that 180 people were enrolled (company statements, no results yet), and a phase 1 trial of a stabilized-stem nanoparticle (H1ssF) registered in January 2026 was withdrawn in September 2026 before enrolling anyone. The one large efficacy trial of a T-cell candidate (MVA-NP+M1, 2,152 adults, 2019) was stopped for futility, with 3.25 percent infected on the vaccine versus 2.14 percent on placebo.
What is missing
- Targets on the virus that stay constant across subtypes and years, and that antibodies can reach
- Vaccines that also build lasting T-cell and airway immunity, not just blood antibodies
- Proof of protection in people, from challenge studies and large field trials
- Better prediction of which strains will circulate, for the yearly vaccine in the meantime
- Stable public funding for large trials
Becomes possible once solved
- One or two shots giving years of protection
- Protection against new pandemic flu strains such as H5N1 before they spread
- No more yearly strain guessing
- Fewer winter hospital admissions
Open steps
- Find the unchanging parts High AI leverageMap which parts of the flu virus, such as the stem of its surface protein, tolerate no change, and which of them antibodies can reach.
- Make conserved parts visible Medium AI leverageDesign vaccine particles that make the immune system respond to hidden conserved regions instead of the variable tip of the surface protein.
- Better yearly strain picks High AI leverageUntil broad vaccines exist, predict months ahead which strains will circulate and which vaccine strain protects best.
- Lasting T-cell and airway immunity Medium AI leverageCreate lasting T-cell and airway immunity that reduces infection, after a large 2019 trial of a T-cell vaccine showed no benefit.
- Prove protection in people Low AI leverageRun human challenge and field trials that show a broad vaccine prevents infection, with agreed endpoints and stable funding.
Where AI could help
Medium AI leverage. AI helps forecast strains and design targets, but protection has to be proven in people through long and costly trials.
- Predicting which strains will dominate next season
- Designing stabilized stem or multi-strain immunogens
- Modelling how mutations let the virus escape antibodies
Shown so far
- VaxSeer, an AI system described in Nature Medicine in 2025, picked H3N2 vaccine strains with a better retrospective antigenic match than WHO's recommendation in 9 of 10 seasons (a surrogate measure, hemagglutinin only). source