{
  "id": 6526454,
  "title": "TAPAS: Learned integration of AlphaFold3 confidence and geometric features for TCR-pMHC binding prediction",
  "url": "https://urgent.news/2026/09/09/tapas-learned-integration-of-alphafold3-confidence-and-geometric",
  "topic": "science",
  "section": "Science",
  "published": "2026-09-09T00:00:00.000Z",
  "source": {
    "name": "bioRxiv",
    "slug": "biorxiv",
    "url": "https://www.biorxiv.org/content/10.64898/2026.09.08.749548v1?rss=1"
  },
  "original_language": "en",
  "account": "Recent advancements in biomolecular structure prediction have led to the development of AlphaFold3, a powerful tool for predicting TCR-pMHC binding specificity. However, individual confidence metrics from AlphaFold3 vary in their predictive performance across datasets, indicating the potential benefits of combining multiple signals for improved accuracy. To address this challenge, researchers have introduced TAPAS, a tabular learning framework that integrates AlphaFold3-derived interface confidence and structural geometry with sequence embeddings. By incorporating these complementary signals, TAPAS aims to overcome the limitations of relying on any single metric for TCR-pMHC binding prediction. Extensive testing on the VDJdb and two external benchmarks revealed that TAPAS consistently outperformed other zero-shot methods, achieving the highest ranking and matching or surpassing the strongest individual AlphaFold3 confidence metric. Further analysis through feature group ablation demonstrated that sequence, confidence, and geometric features each contributed differently depending on the evaluation setting, underscoring the importance of integrating these diverse features within a unified framework. This innovative approach not only highlights the value of combining structural and sequence features but also paves the way for more robust and reliable TCR-pMHC binding prediction.",
  "summary": "Motivation Recent advances in biomolecular structure prediction, exemplified by AlphaFold3, have opened new opportunities for the prediction of TCR-pMHC binding specificity. Although individual AlphaFold3 confidence metrics provide informative binding signals, their predictive performance varies across datasets, highlighting the need to combine complementary signals rather than rely on any single…",
  "key_points": [],
  "editors_take": null,
  "illustration": null,
  "coverage": {
    "outlets": 1,
    "also_reported_by": []
  },
  "ai_generated": true,
  "disclaimer": "Summaries, key points and the editor’s take are written by software from other outlets’ reporting and may contain errors — always check the linked original."
}