On TCR binding predictors failing to generalize to unseen peptides

• DOI: 10.3389/fimmu.2022.1014256
• Publication Date: 2022-10-21T06:55:20Z
• Authors: Filippo Grazioli; Anja Mösch; Pierre Machart; Kai Li; Israa Alqassem; Timothy J. O’Donnell; Martin Renqiang Min
• Keywords: Training set

Several recent studies investigate TCR-peptide/-pMHC binding prediction using machine learning or deep approaches. Many of these methods achieve impressive results on test sets, which include peptide sequences that are also included in the training set. In this work, we how state-of-the-art models for generalize to unseen peptides. We create a dataset including positive samples from IEDB, VDJdb, McPAS-TCR, and MIRA set, as well negative both randomization 10X Genomics assays. name collection TChard . propose hard split , simple heuristic training/test split, ensures exclusively present peptides do not belong effect different splitting techniques models’ performance, testing mismatched generated randomly, addition derived Our show modern fail provide an explanation why happens verify our hypothesis dataset. then conclude robust TCR recognition is still far being solved.