Self-similarity and tolerance
In one sentence
Immune tolerance restrains responses to self, while a candidate target's resemblance to healthy molecules is one factor in recognition and safety.
The intuition
The immune system needs both recognition and restraint. A receptor that can recognize something is not automatically a receptor that should attack it. Several layers of education and regulation reduce harmful responses to healthy tissue.
Sequence comparison is useful, but a molecular “looks familiar” score is not a complete safety inspection.
How it works
During T-cell development, central tolerance removes many cells that react strongly to self targets and supports the development of regulatory populations. Peripheral tolerance restrains mature cells through mechanisms such as limited activation signals, functional unresponsiveness and regulatory T cells. These mechanisms differ from having no receptor capable of binding a self molecule.
A tumor-specific mutation may make a peptide distinguishable from its normal counterpart. That can create an opportunity for recognition, but the altered peptide still needs suitable human leukocyte antigen (HLA) display and a responsive T-cell receptor (TCR). A large sequence change does not guarantee either step.
Conversely, a small change can matter if it alters peptide binding or a receptor-contact surface. There is no universal percentage of sequence identity that turns a candidate into a safe or effective treatment. Receptors can cross-react with different peptides, including peptides from unrelated proteins. Molecular shape and the presenting HLA molecule contribute alongside sequence.
Engineering receptor affinity can change the balance. Stronger binding is not automatically more selective. Evaluation therefore includes the normal counterpart, other plausible cross-reactive peptides and relevant healthy-cell systems, with controls that test the actual recognition route.
Why it matters in cancer
Vaccine selection asks whether a response can distinguish tumor from healthy tissue. Engineered-receptor development also asks whether the introduced receptor recognizes an unintended target. These questions overlap, but a computational comparison and a functional safety experiment provide different evidence.
Worked example
Fictional mutant peptide M differs by one amino acid from normal peptide N. A binding model favors M. In a functional assay, a receptor responds strongly to both M and N when each is experimentally loaded onto matching HLA cells.
The prediction did not establish mutant-specific recognition. The next questions concern naturally processed display, response thresholds and healthy tissues. A second receptor that responds to M but not N in this assay is more discriminating under the tested conditions; it still needs broader cross-reactivity testing.
Common confusions
- Tolerance is a collection of biological mechanisms, rather than a single sequence filter.
- Tumor-specific sequence does not establish tumor-exclusive receptor recognition.
- A stronger receptor can increase unwanted recognition as well as intended recognition.
- A negative result in one healthy-cell model does not cover every tissue or experimental condition.
Related concepts
Sources and scope
Source-checked October 9, 2026. The example is fictional and supplies no clinical selection threshold. Expert and learner review remain pending.
- Janeway's Immunobiology: self-tolerance and its loss, for central and peripheral tolerance mechanisms.
- Cameron and colleagues: a titin-derived cross-reactive target, a primary demonstration that an engineered MAGE-A3-directed TCR recognized another peptide.
- Łuksza and colleagues: neoantigen quality and immunoediting, a primary pancreatic-cancer study modeling similarity and recognition; its model is not a universal clinical safety rule.