The Immune Epitope Database (IEDB) is a freely accessible scientific resource containing curated experimental information about epitopes: molecular structures recognized by antibodies or immune-cell receptors. It brings together findings from published studies and direct data submissions, recording both the structures tested and the experimental contexts in which recognition was assessed. Its associated analysis resource provides computational tools for investigating immune recognition and predicting candidate epitopes. Supported by the United States National Institute of Allergy and Infectious Diseases, the IEDB is maintained through a project led by the La Jolla Institute for Immunology.
Biological scope
An epitope is the part of an antigen involved in a specific recognition event. The IEDB distinguishes several related kinds of experimental evidence.
B-cell and antibody recognition. Antibodies and the antigen receptors of B cells can recognize continuous molecular sequences or discontinuous surfaces formed when an antigen folds. Consequently, an antibody epitope may be represented as a linear sequence, a set of residues brought together in a three-dimensional structure, or a nonpeptidic chemical structure.
T-cell recognition. Conventional T cells recognize peptides presented by major histocompatibility complex molecules. Recognition depends on the peptide, the presenting molecule, and the T-cell receptor. The same peptide can therefore produce different results in different experimental or genetic contexts.
MHC binding and ligand presentation. Experiments can measure peptide binding to MHC molecules or identify peptides naturally associated with them. These observations are relevant to antigen presentation but do not, by themselves, establish that a peptide elicits a T-cell response.
The database covers experiments involving humans and other host species. Its subject matter includes infectious agents, allergens, and self-antigens studied in autoimmunity, among other areas. Epitope structures are not limited to protein fragments; nonpeptidic antigens can also be represented.
Data organization and curation
The IEDB is organized around experimental evidence rather than a simple catalogue of sequences. A single epitope can be associated with many assays, publications, hosts, and outcomes.
A curated record may describe:
- Epitope structure: a sequence, defined residues, chemical identity, or structural description.
- Source antigen and organism: the molecule and biological source from which the tested structure derives.
- Host: the species and relevant characteristics of the individual or experimental animal.
- Exposure history: infection, immunization, or another exposure associated with the immune response.
- Assay method: what was measured and how the experiment was performed.
- Recognition context: antibody identity, responding cell type, or MHC restriction, when reported.
- Outcome: positive or negative findings and quantitative measurements where available.
- Provenance: the publication or submission supporting the observation.
Curators translate information from scientific reports into structured fields and standardized terminology. This makes studies with different wording and reporting styles searchable within a common framework. Curation preserves distinctions among the source organism, the experimental host, and the material used in an assay, which are not necessarily the same.
Negative results are important because they document tested structures that did not meet an assay’s criteria for recognition. However, a negative result applies to the reported conditions; it is not evidence that a structure can never be recognized.
Searching and interpreting records
The search interface allows users to combine criteria such as epitope sequence, source antigen, organism, host, assay type, MHC restriction, and experimental outcome. Results can be examined at different levels, including epitopes, assays, and references, and exported for further analysis.
These levels answer different questions. The number of distinct epitopes is not the number of independent experiments, and the number of positive assays is not the number of responding individuals. A frequently studied peptide may have many records because it has been tested repeatedly.
Sequence searches also require biological interpretation. An identical sequence can occur in different source proteins or organisms. Conversely, related sequences may differ in immune recognition because of a mutation, antigen processing, or differences in the presenting MHC allele. Sequence similarity alone does not demonstrate cross-reactivity.
Analysis and prediction tools
The IEDB Analysis Resource complements the experimental database with computational methods. Its principal functions include:
- MHC class I and class II binding prediction, which estimates how candidate peptides interact with particular MHC molecules.
- Antigen-processing and presentation prediction, which models selected steps affecting whether a peptide may become an MHC-associated ligand.
- B-cell epitope prediction, using sequence-based or structure-based methods to identify candidate antibody-recognition regions.
- Population coverage analysis, which combines specified MHC restrictions with allele-frequency information to estimate potential coverage.
- Epitope conservancy analysis, which measures how widely a sequence is retained across a supplied set of related sequences.
Some methods use machine learning trained on experimental datasets. Their outputs are predictions, not additional experimentally established epitopes. A strong predicted MHC binder need not be naturally processed, presented at sufficient abundance, or recognized by available T cells. Likewise, a predicted antibody-accessible region is not necessarily a functional antibody epitope.
Scores and percentile ranks must be interpreted according to the particular method. Outputs from different models are not automatically interchangeable.
Research uses
The IEDB supports the comparison of immune responses across antigens, organisms, and experimental systems. Researchers use it to locate previously tested epitopes, identify gaps in experimental coverage, assemble peptide sets for immune monitoring, and investigate cross-reactivity.
In vaccine research, curated evidence and prediction tools can help prioritize candidate antigen regions for experimental study. In allergy and autoimmunity research, records help organize evidence about molecular targets of immune recognition. These uses concern research design and evidence synthesis, rather than establishing clinical effectiveness.
The database also supplies training data and evaluation datasets for computational immunology. MHC-binding measurements, naturally presented ligands, and functional T-cell assays represent different prediction targets and must be distinguished when constructing such datasets.
Limitations and methodological considerations
Coverage reflects research activity. Well-studied pathogens, antigens, host populations, and MHC alleles are more extensively represented than less-studied ones. Absence from the database is not evidence that an epitope does not exist.
Published evidence is selective. Publication bias can affect which experiments become available for curation. Reported negative results cannot represent all unsuccessful or unreported experiments.
Assays measure different phenomena. Binding, presentation, cellular activation, and protection against disease are related but distinct outcomes. A cytokine response, for example, is not equivalent to demonstrated protective immunity.
Experimental context matters. Peptide concentration, sample preparation, host exposure history, and detection thresholds can influence results. Standardized annotation improves comparability but does not make heterogeneous experiments identical.
Computational evaluations can contain information leakage. Closely related peptides, repeated measurements, or data from the same study may appear in both training and test sets. Performance estimates can consequently overstate generalization to genuinely unfamiliar antigens or MHC alleles.
The resource changes over time. New curation, corrected annotations, and updated prediction methods can alter search results and analytical outputs. Reproducible research therefore depends on recording retrieval dates, search criteria, dataset versions where available, and the exact methods used.