Vaccine response: a predicted model that will respond poorly, based on antibodies already present

Researchers at Arizona State University measured antibodies against 185 antigens in more than 4,000 people to identify a signature of "immune readiness."

Predicting the intensity of vaccine response prior to injection: this is suggested by a study at Arizona State University, involving 8,687 blood samples from 4,089 participants. The panel included healthy volunteers and people with HIV-related immunosuppression, multiple myeloma, solid tumour, autoimmune disease, inflammatory bowel disease or organ transplant.

For each sample, researchers measured antibodies against 185 antigens — common viruses and bacteria, but also targets associated with autoimmune diseases. A deep learning model then looked for regularities on the entire panel rather than isolated markers, in order to reconstruct an overall picture of the immune state of each participant.

The ranking obtained does not overlap the expected clinical categories: some healthy people respond poorly, while immunocompromised patients develop a strong response. The approach has a practical advantage over genotyping strategies: it requires only a blood sample, which would make it easier to translate into clinical practice.