TITANiAN: Robust Prediction of T-cell Epitope Immunogenicity using Adversarial Domain Adaptation Network

T-cell immunogenicity, the ability of peptide fragments to elicit T-cell responses, is a critical determinant of the safety and efficacy of protein therapeutics and vaccines. While deep learning offers promise for in silico prediction, the scarcity of comprehensive immunogenicity data remains a major challenge. We present TITANiAN, a novel multi-domain deep learning framework that leverages adversarial domain adaptation to integrate diverse immunologically relevant data sources, including MHC presentation, pMHC binding affinity, TCR-pMHC interaction, T-cell activation, and source organism information. Validated through rigorous leakage-controlled benchmarks, TITANiAN demonstrates exceptional performance in its primary fine-tuned task (TITANiAN-IM), predicting T-cell activation for specific peptide-MHC pairs. Remarkably, it also predicts ADA-inducing potential of therapeutic antibodies without requiring MHC inputs. This success attributed to TITANiAN’s biologically grounded and data-driven multi-domain pre-training, with component contributions further confirmed by extensive ablation studies. In addition, TITANiAN supports broader immunogenicity-related tasks such as molecular binding prediction, enabled by its flexible pretraining architecture. TITANiAN’s consistent and robust performance highlights its potential to advance the development of safer and more effective vaccines and protein therapeutics. The model is publicly accessible via a web server: https://galaxy.seoklab.org/design/titanian/ .