Counterfactual Peptide Editing for Causal TCR–pMHC Binding Inference

Sanjar Khudoyberdiev, Arman Bekov · 2026

Read Counterfactual Peptide Editing for Causal TCR–pMHC Binding Inference by Sanjar Khudoyberdiev, Arman Bekov (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2604-13256.txt.

Description

Khudoyberdiev and Bekov tackle shortcut learning in T-cell receptor binding guesses, where models exploit peptide length bias or gene pairing instead of the real binding surface. Their Counterfactual Invariant Prediction framework makes lifelike limited counterfactual peptide edits, keeping invariance at non-anchor positions while raising response at MHC anchor spots. On a curated VDJdb-IEDB benchmark under family-held-out tests, it reaches AUROC 0.831 and consistency 0.724. That is a 39.7 percent cut in shortcut index against the plain baseline.

Catalog record

CREATORSanjar Khudoyberdiev, Arman Bekov
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE20629
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2604.13256

Browse the full library · About the Tower of Babel