Lab AI flags weak vaccine responses before vaccination

In a bank of 8,687 blood samples from 4,089 people, researchers led by Joshua LaBaer at Arizona State University found a way to look for vaccine readiness before the injection. Their artificial-intelligence model used antibody patterns measured before and after COVID-19 vaccination to distinguish people more likely to mount strong responses from those more likely to respond weakly.

The team measured antibodies against 185 antigens—targets recognized by the immune system—including SARS-CoV-2, common viruses and bacteria, and markers associated with autoimmune diseases. Instead of checking only whether a person already carried antibodies against one pathogen, the researchers built a wider picture of the immune system's existing activity.

Some antibodies appeared to serve as "sentinels." Higher levels against microbes including Staphylococcus aureus, RSV and human respirovirus 3 were associated with stronger COVID-19 vaccine responses. They do not necessarily attack the vaccine target directly; the researchers say they may signal how responsive the antibody-producing arm of the immune system is at baseline. The deep-learning model then combined many such measurements to detect patterns that individual markers could miss.

The result matters because broad health categories do not tell the whole story. Several immunosuppressed groups were more likely to have blunted responses, but some participants in those groups responded strongly, while about 5% to 6% of healthy participants had weak responses. A blood-based fingerprint could eventually help clinicians find those people before protection is assumed, rather than discovering the problem only after vaccination.

Concretely, the proposed use is targeted care: identify patients who may need additional vaccine doses, closer follow-up or alternative protective measures. The study's findings could extend beyond COVID-19, but the researchers say that requires validation in additional studies and with other vaccines. The measurements and predictions come from the research team; broader clinical use has not yet been established. The work appears in Cell Press Blue.