A tumour, viral or normal cellular protein exists inside a cell.
Beginner reference
Back to primerConcepts and glossary
Start with the recognition pathway, then use the glossary whenever a term appears elsewhere in TCRForge.
Core biology
How a T cell sees a target
The TCR usually recognizes a peptide and HLA together—not a free antigen by itself.
The protein is processed into short amino-acid fragments.
An HLA molecule holds one peptide and displays it on the cell surface.
The paired TCR reads the combined peptide–HLA surface.
Signalling may trigger activation, cytokine release or killing.
Connection to in vivo CAR-T
Where the TCR-alpha promoter topic fits
The delivery particle, expression switch and receptor payload solve three different problems.
Cell targeting
Engineered vector-envelope binders determine which cells the delivery particle enters; detargeting changes which native receptors it avoids.
This controls entry. A TCR/TRAC promoter does not make a vector enter T cells.Expression gating
A T-cell-selective promoter determines whether the delivered CAR gene is transcribed after the vector has entered a cell.
This is where TCR-alpha/TRAC regulatory biology is directly relevant, although a generic T-cell promoter is not necessarily the TRAC promoter.Receptor payload
The CAR combines an antibody-derived binding domain with hinge, transmembrane, costimulatory and CD3ζ signalling domains.
This is CAR engineering. A natural αβ TCR instead recognizes peptide–HLA and uses the native CD3 complex.Reference index
Terms used across the platform
33
T cellT lymphocyteImmune foundations
An immune cell that surveys other cells and can coordinate or directly kill when it recognizes a relevant target.
Think of it asA mobile inspector with an identity reader on its surface.Its activation state, source and manufacturing history can change the behavior of the same engineered receptor.
Experiment context and decision reviewAntigenImmune foundations
A biological molecule that can be recognized by the immune system. For T cells, the relevant target is usually a peptide cut from a protein.
Think of it asThe source document from which a short identifying excerpt is taken.A whole protein name is not enough to define a TCR target; the exact peptide and HLA are required.
Receptor intakePeptideImmune foundations
A short chain of amino acids, commonly 8–15 residues for many HLA-presented T-cell targets.
Think of it asA short excerpt displayed from a much larger protein.Changing one amino acid can change HLA loading, TCR recognition or cross-reactivity.
Receptor evidence planHLAHuman leukocyte antigenImmune foundations
A human cell-surface protein that holds a peptide and displays it for T-cell inspection. HLA is the human form of MHC.
Think of it asA display tray plus an identity frame: the TCR reads the peptide and the tray together.Different HLA alleles present different peptide sets. A TCR result tied to HLA-A*02:01 does not automatically transfer to another allele.
Receptor intake, model inputs and assay contextpMHCPeptide–MHC complexImmune foundations
The combined molecular surface formed when a peptide sits in an HLA/MHC binding groove.
Think of it asThe message and the display device are one recognition object.Most TCRs recognize the geometry of peptide plus HLA, not a free peptide alone.
Structure models and specificity evidenceTCRT-cell receptorImmune foundations
The receptor on a T cell that recognizes a particular peptide–HLA surface and initiates signaling.
Think of it asA molecular reader whose shape and chemistry define what it can notice.Sequence alone is not the whole product: pairing, expression, cell state and cross-reactivity all matter.
Receptors, constructs and model registryTCR α/β chainsImmune foundations
The two protein chains that pair to form the common αβ TCR recognition surface.
Think of it asTwo halves of one reader; evaluating only one half loses context.Many datasets contain only the beta chain, so their predictions have a built-in information ceiling.
Receptor intake and model evidence tierCDR3Complementarity-determining region 3Immune foundations
The most variable loop in each TCR chain and a major contributor to peptide–HLA recognition.
Think of it asThe reader’s most customized contact edge.CDR3 is useful but insufficient by itself; V/J genes, the paired chain and HLA context also contribute.
Sequence inputs and specificity modelsClonotypeImmune foundations
A distinct T-cell receptor sequence identity, usually defined by its rearranged receptor chains.
Think of it asA receptor barcode shared by descendants of one T-cell clone.Clonotype frequency can suggest immune expansion but does not prove target specificity.
Evidence retrieval and repertoire analysisGene locusLocus and regulation
A physical region of the genome containing a gene and the surrounding regulatory DNA that controls it.
Think of it asNot only the sentence, but the page layout, headings and nearby instructions.Moving a sequence to a new locus can change its timing, level and cell-state dependence.
Locus atlasTRA/TRD locusLocus and regulation
The chromosome 14 region containing TCR alpha genes and the nested TCR delta locus.
Think of it asA rearranging genomic workspace shared by two receptor programs.Rearrangement and cell maturation mean the engineered allele cannot be understood as a simple static promoter cassette.
Locus atlas and construct comparisonTRACT-cell receptor alpha constantLocus and regulation
The constant-region gene used by the TCR alpha chain and a common target for orthotopic TCR knock-in.
Think of it asA native receptor anchor that can be reused to place an engineered receptor under endogenous control.Targeting TRAC may reduce native receptor competition, but the exact donor and junction design still require evidence.
Locus atlas, constructs and decision gatePromoterLocus and regulation
A DNA region near a transcription start site that recruits the machinery needed to begin RNA production.
Think of it asThe local start switch for reading a gene.Promoter behavior depends on sequence, chromatin, enhancer contacts and genomic location.
Locus evidence and construct controlEnhancerLocus and regulation
A regulatory DNA element that can increase or shape gene expression, often across a long genomic distance.
Think of it asA remote control that changes how a nearby start switch behaves.An enhancer is not a portable promoter; its effect can depend on topology and cell state.
Locus atlas and regulatory evidenceEndogenous regulationLocus and regulation
Expression control supplied by the cell’s native genomic locus and regulatory network.
Think of it asUsing the cell’s existing control system instead of installing a separate always-on switch.It may improve physiological timing and expression, but must be validated after the exact edit.
Construct architecture comparisonCRISPR–Cas9Cell engineering
A programmable genome-editing system that uses a guide RNA to direct Cas9 to a selected DNA sequence.
Think of it asA GPS-guided molecular cutter.Cut location, repair pathway and delivery method jointly determine the resulting cell population.
Construct and genome-safety evidenceKnock-inCell engineering
The intentional insertion or replacement of DNA at a chosen genomic site.
Think of it asInstalling a new module at a specified address.The intended insertion can coexist with partial, incorrect or structurally altered outcomes.
Construct comparison and genome QCKnockoutCell engineering
A genome edit intended to disable a gene or prevent it from producing a functional product.
Think of it asDisconnecting an existing module.Disrupting native TCR genes may reduce receptor competition, but editing outcomes and cell effects require measurement.
Construct architecture and safety reviewDonor templateCell engineering
DNA supplied to a cell as the sequence blueprint for a targeted insertion or repair.
Think of it asThe replacement part and installation instructions delivered together.Payload geometry, homology arms and junctions affect integration efficiency and genomic integrity.
Construct design and experiment matrixOrthotopic integrationCell engineering
Placing an engineered gene into the native or functionally corresponding genomic locus.
Think of it asInstalling the replacement in the original control socket.It can preserve native regulation better than a separate cassette, but that benefit is architecture- and context-dependent.
TRAC architecture comparisonOn-target / off-targetCell engineering
On-target describes events at the intended genomic site; off-target describes unintended editing elsewhere.
Think of it asCorrect address versus unintended addresses.A clean off-target screen does not rule out complex structural damage at the intended site.
Genome-safety evidence and decision blockersConstructCell engineering
The designed DNA and protein architecture used to build an engineered cell product.
Think of it asThe full wiring diagram, not just one component.Receptor sequence, control elements, linkers, editing strategy and locus must be evaluated together.
Construct architecture comparisonSpecificityEvidence and AI
Which molecular target a receptor recognizes relative to alternative targets.
Think of it asWhether a key selects the intended lock and avoids similar locks.A high predicted score for one target does not establish absence of recognition across the self-peptide universe.
Specificity models, evidence and counter-screensAffinity / avidityEvidence and AI
Affinity is the strength of one molecular interaction; avidity is the combined functional strength of many interactions in a cellular setting.
Think of it asOne hook’s grip versus the total hold of many hooks.Neither quantity alone guarantees useful T-cell function or safety.
Evidence interpretation and functional assaysCross-reactivityEvidence and AI
Recognition of unintended peptides or peptide–HLA complexes by the same receptor.
Think of it asA reader mistakenly accepts a similar-looking identity card.It is a central safety risk and requires counter-screening plus experimental confirmation.
Evidence blockers and experiment matrixPrimary human T cellsEvidence and AI
T cells freshly isolated from human donors rather than an immortalized laboratory cell line.
Think of it asTesting in the intended biological material rather than a simplified stand-in.Donor variability and activation history are real parts of the product context.
Experiment contextFunctional assayEvidence and AI
An experiment that measures what engineered cells actually do, such as activation, cytokine release or killing.
Think of it asA road test rather than an inspection of parts.Binding or structure predictions cannot replace matched functional testing.
Experiment matrix and decision gateEmbeddingEvidence and AI
A learned numerical representation that places sequences or structures with related patterns near one another in a mathematical space.
Think of it asA coordinate system for biological similarity.Embeddings are useful for retrieval and clustering but proximity is not proof of the same specificity.
Model registry and retrieval stageStructure predictionEvidence and AI
An AI or physics-assisted estimate of the three-dimensional arrangement of molecules from sequence and other inputs.
Think of it asA plausible 3D hypothesis, not a photograph.Confidence in geometry does not equal affinity, specificity, signaling or cellular function.
Model registry and Modal runsModel confidenceEvidence and AI
A model-derived estimate of certainty about its own output, often specific to one task or component.
Think of it asHow sure the model says it is—not how true the answer is.Confidence must be calibrated on relevant data before it can support thresholds or decisions.
Model details and run provenanceCalibration / OODOut-of-distributionEvidence and AI
Calibration asks whether predicted probabilities match observed frequencies; OOD describes inputs unlike the model’s training distribution.
Think of it asA trustworthy weather forecast, plus knowing when you have left the mapped climate zone.A high score on familiar data may fail for a new peptide, HLA, donor or experimental protocol.
Model governance and admission gatesNegative sampleEvidence and AI
A labeled example intended to represent a non-binding or non-matching TCR–target pair during training or evaluation.
Think of it asExamples that teach the model what “not the target” means.Randomly generated negatives can make a model look strong while failing against realistic near-miss peptides.
Model AI logic and governanceModel evidence vs experimentEvidence and AI
Model evidence is a computed hypothesis or prior; experimental evidence is an observation produced under defined biological conditions.
Think of it asA forecast versus a measured outcome.TCRForge keeps the two evidence types separate so a confident model cannot silently become proof.
Evidence objects and decision review