henhaobio
Sign in

bioRxivBioinformaticsPreprint

NetTCR: sequence-based prediction of TCR binding to peptide-MHC complexes using convolutional neural networks

Vanessa Isabell JurtziD, Leon Eyrich JesseniD, Amalie Kai BentzeniD, Martin Closter Jespersen, Swapnil MahajaniD, Randi VitaiD, Kamilla Kjærgaard Jensen, Paolo MarcatiliiD, Sine Reker HadrupiD, Bjoern PetersiD, Morten NielseniD

PDFDOI
Structured summary available · Jump to summary ↓

Abstract

Predicting epitopes recognized by cytotoxic T cells has been a long standing challenge within the field of immuno- and bioinformatics. While reliable predictions of peptide binding are available for most Major Histocompatibility Complex class I (MHCI) alleles, prediction models of T cell receptor (TCR) interactions with MHC class I-peptide complexes remain poor due to the limited amount of available training data. Recent next generation sequencing projects have however generated a considerable amount of data relating TCR sequences with their cognate HLA-peptide complex target. Here, we utilize such data to train a sequence-based predictor of the interaction between TCRs and peptides presented by the most common human MHCI allele, HLA-A*02:01. Our model is based on convolutional neural networks, which are especially designed to meet the challenges posed by the large length variations of TCRs. We show that such a sequence-based model allows for the identification of TCRs binding a given cognate peptide-MHC target out of a large pool of non-binding TCRs.

Abstract summary

Claude Opus 5.5 · xhigh effort · 17 Oct 2026 · from the abstract · CC BY-NC-ND 4.0

A sequence-only convolutional network trained on T cell receptor (TCR)-peptide pairs for HLA-A*02:01 can single out TCRs that bind a target peptide among many non-binding ones.

Peptide binding to major histocompatibility complex class I (MHCI) molecules can be predicted reliably for most alleles, but models of TCR recognition of peptide-MHC complexes have stayed weak for lack of training data. NetTCR draws on newly available sequencing data that link TCRs to their cognate HLA-peptide targets, using a convolutional neural network (CNN) built to cope with variable TCR lengths.

Key results

  1. ClaimedTrained on TCR-peptide pairs restricted to HLA-A*02:01, the CNN model can identify the TCRs that bind a given peptide-MHC target within a large pool of non-binding TCRs.

    Only the abstract is available: it states this ability but gives no performance metric, number of peptides or TCRs, or comparison with other methods. · Abstract

Methods, data and reproducibility

Sequence-based predictor of binding between TCRs and peptides presented by HLA-A*02:01, the most common human MHCI allele, built from convolutional neural networks designed for the large length variation of TCRs and trained on sequencing data pairing TCRs with their cognate HLA-peptide targets.

  • TCR sequences paired with cognate HLA-A*02:01-peptide targets from recent next-generation sequencing projects · human
Methods & tools
NetTCR, convolutional neural network
Reproducibility gaps
Only the abstract is available: no code or data links, dataset sizes, performance metrics or software details are given.
Limitations (2)
  • Abstract only: no performance metrics, dataset sizes or baseline comparisons are reported in the available text. · noted in this summary
  • Covers a single MHC class I allele (HLA-A*02:01); whether the approach extends to other alleles is not addressed in the abstract. · noted in this summary

Read this if you work on T cell epitope or TCR specificity prediction and want a sequence-only CNN approach for peptides presented by HLA-A*02:01.

Citations
118

Reproducibility ledger

Nobody has reported trying to reproduce this paper yet. Did you replicate an experiment or run the code? Your report saves the next lab a week.

Sign in to file a reproducibility report.

Related papers

Similarity Map
SelectedSame fieldCross-field
THISLuu '21Bradley '22Jensen '24Jiang '23Lang '26Jiang '24
82% semantic matchBioinformatics
Alan Luu, Jacob R. Leistico, Timothy J. Miller+2 more·Genes·Apr 15, 2021· 38 citations