T-SCAPE: T-cell immunogenicity scoring via cross-domain aided predictive engine

T-cell immunogenicity is a critical determinant of safety and efficacy for
protein therapeutics and vaccines, but prediction is hampered by data
scarcity. We present T-SCAPE, a multi-domain deep learning framework that
uses adversarial domain adaptation to integrate diverse immunologically
relevant data sources, including MHC presentation, peptide-MHC binding
affinity, TCR-pMHC interaction, source organism information, and T-cell
activation. Validated through rigorous leakage-controlled benchmarks,
T-SCAPE shows exceptional performance in predicting T-cell activation for
specific peptide-MHC pairs. Remarkably, it also accurately predicts the
anti-drug antibody-inducing potential of therapeutic antibodies without
MHC inputs, a success attributed to its biologically grounded pretraining.
Confirmed by extensive case studies and ablation studies, T-SCAPE’s
flexible architecture also supports broader tasks like molecular binding
prediction. Its robust performance highlights its potential to advance the
development of safer and more effective biologics.