Add iTCep peptide:TCR binding predictor wrapper

[iskandr]

Wrap iTCep as an mhctools peptide:TCR binding predictor (pMHC_TCR_binding).

Upstream: https://github.com/kbvstmd/iTCep — a fusion neural network for peptide–TCR binding.

Pretrained weights: shipped in-repo at models/iTCep.h5 (~35 MB). No retraining needed.

Framework / deps: TensorFlow/Keras 2.4 — an older TF, cleanest to isolate in its own env.

License: AGPL-3.0 — the strongest copyleft here (its network/server-use clause triggers source disclosure). Definitely do not bundle/redistribute; use the "user brings their own clone" pattern out-of-process (as mhctools/tulip.py does for GPL TULIP). Keep the AGPL boundary clean.

Inputs / output: peptide + CDR3β → binding probability (mhc_dependence="none").

Ease: Easy to run — bundled .h5 + a ready predict.py; only friction is the TF 2.4 pin.

Suggested wrapping approach: subprocess into the user's iTCep checkout in an isolated TF 2.4 env (mirroring the TULIP sidecar), never importing iTCep's AGPL code into mhctools.

---

Surfaced from a survey of TCREpitopeBenchmark for methods with public pretrained weights that run easily for inference.

https://claude.ai/code/session_01LZahFhBSCiehXTESCYQ7wG

[iskandr]

⚠️ Read before investing effort: the generalization ceiling

Every TCR–epitope binding predictor in this backlog — this one included — shares a well-documented limitation: they do only modestly on epitopes seen in training and drop to near-random on unseen/novel epitopes. Independent benchmarks find that no current method reliably generalizes; the "best" scores on unseen data sit around the random floor (e.g. mean AUPRC ~0.52–0.53). A tool can look decent on a test set dominated by common epitopes (GILGFVFTL, NLVPMVATV, …) purely because those TCRs saturate the public databases, then fail on anything new.

References

• Moris et al. 2021, Briefings in Bioinformatics — Current challenges for unseen-epitope TCR interaction prediction (foundational: the unseen-epitope challenge, negative-data bias, imbalanced-epitope distribution)
• Meysman et al. 2023, ImmunoInformatics — Benchmarking solutions to the TCR epitope prediction problem: IMMREP22 workshop report (first public benchmark, 23 models) and the IMMREP23 follow-up
• Nature Methods 2025 — Assessment of computational methods in predicting TCR–epitope binding recognition

Implication for this wrapper: worthwhile for reproducibility, cross-method benchmarking, and the well-characterized "seen-epitope" regime — not as a trustworthy predictor for novel epitopes / neoantigens. When we build it, the wrapper's docstring should state this ceiling so users don't over-trust the output.