Pre-vaccine antibody fingerprints may forecast who responds strongly to vaccination
A person’s immune system may leave clues long before a vaccine is given. In a study led by Arizona State University, researchers used artificial intelligence to analyze pre-vaccination antibody patterns and found signals that separated stronger vaccine responders from weaker ones after COVID-19 vaccination.
The work rests on a deceptively simple idea: not all immune systems are starting from the same place. Prior infections, chronic illness, immune suppression, and even the background mix of circulating antibodies can shape how well a vaccine takes hold. In this cohort, those differences were visible in blood samples drawn before vaccination, suggesting a measurable form of immune readiness that may help explain why some people mount robust protection while others do not.
The study examined 8,687 samples from 4,089 participants and measured antibodies against 185 antigens. Those targets included common viruses and bacteria, SARS-CoV-2, and antigens linked to autoimmune disease. The population was broad enough to capture real-world immune variation: healthy volunteers were included alongside people with HIV, multiple myeloma, solid organ malignancy, autoimmune disease, inflammatory bowel disease, and solid organ transplantation. Several immunosuppressed groups were more likely to have reduced responses to COVID-19 vaccination, yet the signal was not as clean as a simple health label might suggest.
Some immunosuppressed participants still mounted strong responses. A small but clinically important fraction of healthy participants did not.
That result is the point at which conventional triage starts to break down. A clinician looking only at diagnosis or treatment category would miss both the unexpectedly robust responder and the ostensibly healthy person whose immune response lagged behind. Vaccine responsiveness, in other words, appears to be a property of the individual immune system, not just the disease category attached to the chart.
Sentinel antibodies and the biology of a prepared immune system
Certain pre-existing antibodies stood out in the analysis. Higher levels of antibodies against common microbes such as Staphylococcus aureus, RSV, and human respirovirus 3 were associated with stronger responses to COVID-19 vaccination. The investigators describe these signals as “sentinel” antibodies because they seem less like direct drivers of vaccine response and more like markers of whether the antibody-producing arm of immunity is primed to react.
That concept fits immunology. Antibodies reflecting prior microbial exposure may be acting as a readout of immune system experience, reserve, and coordination. A person whose humoral immune system has maintained a broad, responsive memory pool may have a better capacity to expand antibody production after vaccination. In contrast, a flatter pre-vaccine antibody profile could mark immune exhaustion, immune suppression, or simply a less reactive baseline state.
The study did not stop at one or two markers. Investigators used a deep learning model to examine the full antibody fingerprint and search for patterns across the entire panel. That approach matters because immune function is rarely governed by a single biomarker. Antibody responses are shaped by a network of prior exposures, host factors, and immune regulation that conventional linear analyses can miss.
Why health status alone failed as a predictor
The practical challenge is that broad health categories often obscure meaningful biologic diversity. Labels such as “healthy” or “immunosuppressed” are useful shorthand, but they do not fully capture how an immune system behaves on the day a vaccine is administered. Within the immunosuppressed groups, response strength varied. Within the healthy group, a subset still produced weak responses.
About 5% to 6% of healthy participants showed weak vaccine responses. That small percentage carries disproportionate clinical weight because it represents people who would not be flagged by ordinary risk stratification, yet may remain vulnerable after vaccination.
For patients with immune-compromising conditions, the stakes are even higher. If a clinician can identify blunted vaccine response before immunization, the conversation shifts from reassurance to planning. Additional doses, closer post-vaccine monitoring, and other preventive strategies may become more relevant for people who are unlikely to achieve strong protection from a standard schedule. The study does not provide a ready-made clinic tool, but it does define a biologically plausible route toward one.
AI as a search tool for hidden immune structure
The appeal of artificial intelligence in this setting is not mystique; it is pattern recognition at scale. Millions of relationships can exist across antibody signatures, and only a fraction are visible through conventional analysis. Machine learning can search across those relationships simultaneously and detect combinations of antibodies that behave like a higher-order immune fingerprint.
In this study, the models were not just classifying people by diagnosis. They were identifying a pre-existing immune state that predicted later vaccine performance. That distinction is important because it points toward a more individualized form of vaccinology, one that treats the immune system as a dynamic biologic network rather than a single yes-or-no test result.
A broader implication emerges from the biology itself. If sentinel antibodies consistently reflect immune readiness, then pre-vaccine blood profiling could become a way to identify patients who need a different protection strategy before they encounter preventable infection. The framework could also inform vaccine development, helping researchers understand why the same vaccine produces such different outcomes across patients with similar clinical labels.
What this could mean for future vaccination strategies
The investigators suggest that the approach may eventually extend beyond COVID-19 if future studies confirm the findings across other vaccines. That would open the door to a more tailored model of vaccination, especially for patients with impaired immunity, transplant recipients, cancer patients, and others at risk for weak responses.
The immediate clinical attraction is straightforward: predict under-response before protection fails. A blood-based antibody fingerprint would be easier to translate into practice than many genetic assays already used in research settings, and it could be incorporated into existing laboratory workflows if the signal proves stable and reproducible.
Still, the findings remain a first step rather than a finished clinical test. The analysis focused on COVID-19 vaccination, and the path from immune signature to bedside decision-making will require confirmation in additional cohorts and with additional vaccines. The model also needs to show that prediction can improve real-world decisions, not just retrospective classification.
For now, the study offers a clear biologic message. Pre-vaccine immunity is not blank space. It carries information, and in this dataset, that information was strong enough for AI to separate people likely to respond well from those more likely to mount a blunted response.