EMBER predicts TCR-pMHC binding without a docked complex
Original title: EMBER: Predicting TCR--pMHC Binding Without a Docked Complex
Under epitope-disjoint tests it led six retrained baselines on every metric, by 0.9 to 14.4% in micro-AUROC.
Researchers propose EMBER, a structure-based framework that predicts TCR-pMHC binding without a docked complex, avoiding placement error from co-folding. It folds the TCR and pMHC independently, encodes them as SE(3)-invariant residue graphs, and compares representations without a shared coordinate frame. Preprocessing uses O(N+M) monomer folds rather than O(NM) complex operations. Across four folding tools and six retrained baselines, EMBER ranked first on every reported metric.
- Gains were 18.5 to 25.3% in micro-AUPRC and 10.6 to 24.1% in MCC.
- A relative pose is not an input, so placement error cannot propagate.
- Ranking was up to 85% more stable across folding tools.
- Evaluation used epitope-disjoint splits against six retrained baselines.
From the original article
1 min read at huggingface.co
We propose EMBER, a structure-based framework for TCR--pMHC binding prediction that eliminates the need for a predicted docked complex, and with it the placement error that dominates such predictions. Traditional structure-based predictors utilize the co-folded complex structure of the receptor and the antigen and interpret binding from its contact surface.
By Lin Ha, Chen Zhou, Haozheng Luo · huggingface.co
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.