tcren: structure, recognition and ranking#
TCRen provides two routes from TCR–pMHC structures to experimental comparisons:
Statistical contact potentials: rank candidate epitopes and relate structural readouts to measured affinity, response or kinetic endpoints. Start with Task-specific ranking scores; use Evaluation and reproduction commands for matched experimental comparisons.
Confidence, topology and enrichment: combine ipTM, recognition graph cycles and mock-calibrated contact enrichment to rank candidate receptors against one fixed pMHC. Evaluate supplied binder/nonbinder or binder/mock labels on the declared population. Start with Install and rank models and Screen a cohort.
The inspection toolbox exports contact chemistry, residue contributions, geometry and pMHC surfaces. Use Working with modelled complexes to interpret a model, Interface feature reference to choose an export, and Figure gallery for molecular views.
$ tcren screen -s models.tar.gz --confidence confidence.tsv -o scores.tsv
The default TCR score’s weights depend on the unlabelled cohort mean pLDDT. Ranking scores are dimensionless; experimental associations do not calibrate them into binding probabilities, affinities or kinetic rates.
Learn with a worked example#
Complete a task#
Understand the result#
Look up definitions and APIs#
- Screen a cohort
- MHC annotation and supplied alleles
- Interface feature reference
- Every descriptor, with its units
- Potentials
- Experimental contact penalties
- Generated-model inspection
- pMHC feature–performance associations
- Evaluation and reproduction commands
- API reference
- CDR3 chemical-neighborhood diagnostics
- Within-residue atom-pattern potential — frozen design, 21 October 2026
- Inventory-aware whole contact patterns
- Chemical core: frozen numerical specification
- Experimental ranking APIs
- Structure summary (oracle)
- Assessing a modelled complex
- Historical contact-map Potts model
- Reliability: scoring one modelled structure