The immune system’s response to a vaccine is the product of a complex, highly individualized set of biological processes, shaped by a person’s genetics, health history, and years of prior exposure to microbes. That complexity helps explain why vaccines, while reliably protective for the vast majority of recipients, do not produce perfectly uniform results, and why clinicians have had few tools to identify in advance which patients might need closer follow-up. A new study suggests that answer may already be detectable in a person’s bloodstream before they ever roll up their sleeve, opening a path toward more tailored vaccination strategies for those most likely to benefit from them.
The research, published in the journal Cell Press Blue, comes from researchers at Arizona State University’s Biodesign Institute, working with collaborators at Icahn School of Medicine at Mount Sinai, the Feinstein Institutes for Medical Research, the University of Minnesota, and other institutions. The study was conducted through the National Cancer Institute’s Serological Sciences Network, established to evaluate COVID-19 vaccine performance across populations with weakened immune systems.
Sentinel Antibodies Emerge
The team drew on more than 8,600 blood samples from 4,089 people, collected before and after COVID-19 vaccination. Roughly half the participants were healthy volunteers, while the rest had conditions or treatments known to suppress immune function, including HIV, multiple myeloma, solid organ transplants, autoimmune disease, and inflammatory bowel disease. Using a high-throughput lab technique, researchers measured each person’s existing antibody levels against 185 different targets, spanning common viruses and bacteria as well as markers associated with autoimmune conditions, well before anyone received a vaccine dose.
As expected, several immunosuppressed groups were more likely to mount a weak antibody response to vaccination, particularly solid organ transplant recipients. But the immunosuppression categories themselves proved to be imprecise predictors: some transplant recipients responded robustly, while roughly 5 to 6 percent of ostensibly healthy participants turned out to be weak responders despite having no underlying condition at all. That overlap is what pushed the team to look for a better signal.
People who already carried higher levels of antibodies against certain common, unrelated microbes, including Staphylococcus aureus, respiratory syncytial virus (RSV), and human respirovirus 3, tended to mount stronger antibody responses to the COVID-19 vaccine later on, a pattern that held across both healthy and immunosuppressed groups. The researchers call these “sentinel” antibodies. They are not thought to interact with the vaccine target directly. Instead, their presence appears to signal that a person’s broader antibody-producing machinery is primed and functioning well. When the team fed a wide profile of these antibody patterns, rather than just a handful of markers, into a deep learning model, it could distinguish likely strong responders from likely weak ones with reasonable accuracy, including among healthy individuals with no obvious risk factors.
Predicting Who Needs a Backup Plan
If the findings hold up in further studies, they point toward a way to flag vaccine non-responders before a shot is even given, using a standard blood draw rather than genetic sequencing, which the study’s authors note would be considerably harder to deploy in routine clinical settings. That could allow doctors to identify people, inside or outside recognized risk groups, who might benefit from an additional dose, closer monitoring after vaccination, or an alternate form of protection.
The approach also reframes how researchers think about immune readiness. Rather than asking whether someone has previously encountered a specific pathogen, the sentinel-antibody concept treats the immune system’s response capacity as a general trait that can be read off a broad antibody fingerprint, largely independent of the specific vaccine target. For biosecurity and pandemic preparedness planning, a validated version of this approach could help identify vulnerable populations ahead of a rollout, rather than discovering gaps in protection only after an outbreak exposes them.
The authors caution that their models performed well in healthy, HIV, and multiple myeloma cohorts but performed poorly in others, notably autoimmune disease and transplant patients, groups where predicting vaccine response may depend on additional factors such as specific medications or treatment history that the study did not fully capture. The findings are also drawn entirely from COVID-19 vaccination; whether the same sentinel antibodies predict responses to other vaccines remains to be formally tested, though the researchers note a preliminary signal involving measles antibody levels that hints the pattern may extend beyond a single vaccine.
Editor’s Note: This study was made possible by a robust public research infrastructure: the National Cancer Institute’s Serological Sciences Network, established under the National Institutes of Health, part of the Department of Health and Human Services. Coordinating blood samples and antibody data from more than 4,000 participants across multiple institutions required years of sustained federal investment, the kind of long-horizon, unglamorous infrastructure that rarely makes headlines until it produces a finding like this one. Federal science agencies including NIH, NCI, and CDC have faced significant funding uncertainty and workforce reductions in 2025 and 2026, developments Global Biodefense has covered extensively. Findings like this one are a reminder of what that infrastructure makes possible, and what is lost when it is not sustained.
Sources and further reading
Song L, et al. Pre-vaccine sentinel antibodies predict blunted vaccine responses. Cell Press Blue, August 20, 2026.
Why immune responses to vaccines vary from person to person — Arizona State University
This article was researched and sourced by Global Biodefense editors and reported with Claude AI assistance for drafting and editing.