The immune system is central to human health. When it fails, people can become more susceptible to infection and cancer. When misdirected, it can attack the body’s own tissues. And as it ages, it can contribute to chronic disease.

People of the same age can have strikingly distinct immune systems, shaped by genetics and by environmental and lifestyle factors, including air pollution, smoking, diet and past infections^1,2. Growing evidence suggests that variation in the immune system’s ‘biological age’ — the rate at which it has aged, which is distinct from chronological age — matters for health. This variation has been linked to neurodegeneration, cancer and mortality risk³.

Immune cells have a surprising role in exercise endurance

Yet, clinicians have no routine way of measuring how healthy someone’s immune system is. That’s because the immune system is a dynamic network that can change over time. It involves specialized organs that support immune-cell development and function, as well as immune cells that can circulate throughout the body and reside in tissues⁴. Signalling molecules, such as cytokines and chemokines, allow these immune cells and tissues to communicate.

Clinical tests provide only a fragmented picture of the immune system. A physician might examine an individual’s white blood cell count to establish whether circulating immune cells are at unusually high or low levels. They might also measure blood biomarkers such as C-reactive protein, which can indicate inflammation. But neither measurement is specific to a particular health condition, and they do not indicate whether the immune system as a whole is resilient or in decline, or if it is capable of mounting an effective response to infections and other challenges.

But advances in immunology, molecular profiling, medical imaging and artificial intelligence are opening up fresh possibilities for measuring the state of people’s immune systems. The stage is set for researchers to develop the immunome: a comprehensive characterization of immune health across cells, tissues, organs and molecular interactions (see ‘Defining the immunome’).

Research initiatives such as the Human Immunome Project, which one of us (J.S.T.) co-leads, are beginning to provide the data needed. What remains unclear is whether this information could be used to improve patient care, and how that could be achieved. Which measurements best capture clinically important aspects of a person’s immune health? How stable are these variables over time? And, most importantly, can assessing them improve clinical decision-making and, ultimately, disease outcomes?

Measuring every aspect of the immune system will not be feasible for every person. Instead, researchers should try to identify key features that can be readily assessed in the clinic to flag aspects of the immune system that are weakened, or declining quickly when they shouldn’t be. Such assessments could open avenues for disease prevention, treatment guidance and methods to monitor interventions.

For instance, these features might enable physicians to recognize cases of accelerated immune ageing and spot individuals who need close disease monitoring. For cancer treatment, measures of immune health could help to identify people who are most likely to benefit from therapies that rely on the immune system^5,6, as well as those with a high risk of adverse reactions.

Such measurements could also help researchers to understand why people of similar ages and apparently similar health sometimes have substantially different responses to the same infection — a phenomenon that was apparent during the COVID-19 pandemic. Antibody responses to the same vaccine likewise vary widely between individuals^2,7. If this variability could be measured, clinicians could use it to tailor preventive measures, treatments and monitoring to an individual’s immune system.

Researchers, health-care systems and funders must work to find the smallest, most scalable set of measurements that can capture meaningful differences in immune function and guide clinical decisions. Here, we outline a road map for moving the immunome from the laboratory to the clinic.

Deep research

First, scientists should generate longitudinal data sets using large research populations and biobanks. Gaining a full understanding of immune health will require measuring several components of the immune system alongside broader indicators of health and disease in the same people over time.

Analysis of genes and gene expression in single cells from large research populations and biobanks could help researchers to understand which immune-cell groups are present in distinct populations. This could reveal which genes and signalling pathways are active in particular populations of cells, and what combination of markers best distinguishes between healthy, declining and dysfunctional immune states^2,3,8. Analysing blood proteins can indicate whether a person is experiencing inflammation and whether their immune system has been activated. And studying the receptors on immune cells called lymphocytes can reveal the diversity of cells available to recognize threats. These receptors determine which pathogenic antigens the cells will identify; greater diversity generally allows the immune system to detect a broader range of threats.

The Moderna cancer vaccine offers hope — now we must speed up personalized therapies

Researchers must also study how the immune system changes over time to respond to challenges. Studying people before and after they are vaccinated is one standardized way to do this. It’s already clear that a person’s immune state before vaccination can predict the strength and characteristics of the immune response that follows^2,9. The task now is to find out whether a person’s response to routine vaccination could provide a practical read-out of their immune competence¹⁰, and whether it would be possible to modify the immune health of an individual before interventions such as vaccination or cancer immunotherapy to improve their response^2,5.

Data sets used to probe the immunome must represent the diversity of human immunity, because genetics and environmental factors alter immune states¹. For example, malaria vaccine efficacy can differ markedly between populations across continents^7,11. Yet, many existing immune-profiling data sets are concentrated in populations from high-income nations. When studying diverse populations, researchers around the world should use the same protocols so that results from various regions can be compared.

A big challenge will be making sense of complex data sets. AI tools can identify patterns across thousands of immune measurements, but those patterns might reflect differences between populations, laboratories or technologies rather than underlying immune health. Models will therefore need to be tested across independent populations and over time to establish which signals are robust and clinically meaningful.

Scientists working on the Human Immunome Project are developing and applying AI approaches to immunome data collected from various populations. An important goal for the project is to create reference maps and foundation models that scientists can use for tasks such as identifying small, manageable sets of measurements that would be useful in clinical settings.

Find what works

With large-scale data in hand, researchers should work towards simplification. The success of the field should be judged not by how many immune variables can be measured, but by whether it is possible to improve clinical decision-making with just a small, manageable set of them. Tests used in the clinic would ideally be cheap and should not put extra demands on patients. Before any AI tools can be used routinely, they will need to be tested across a range of hospitals and populations. Researchers must establish whether a tool’s outputs truly improve treatment decisions.

One example of a tool that might be useful in the clinic is an AI model capable of extracting information about immune health from routine blood tests and clinical information such as age, sex, body weight and medical history. In a study published last year⁶, such an approach was able to predict health outcomes after immunotherapy treatment for people with cancer.