Unveiling immune targets via TriStageHLA: high-throughput screening of peptide presentation, peptide-HLA binding, and immunogenicity
- 1City University of Hong Kong
Abstract
High-throughput screening of HLA class I epitopes underpins neoantigen vaccine design, T cell therapies, and pathogen immune monitoring, yet proteome-scale prioritization remains challenging under severe class imbalance. We present TriStageHLA, a compact three-stage framework that decomposes epitope discovery into peptide-only screening (PRE), allele-specific binding/presentation prediction (BIND), and immunogenicity prediction (IM). PRE performs early pruning via a LoRA-adapted ESM-2 classifier; BIND integrates quantitative binding affinity and eluted ligand data with a dual-branch architecture combining global cross-attention and local motif extraction; IM is fine-tuned on curated human T-cell assay data with a decoupled LoRA head. Trained on IEDB-derived datasets with per-allele balancing and allele-discordant supervision, TriStageHLA-BIND achieves AUROC of 96.2% and AUPRC of 95.2% on combined BA/EL evaluation, while TriStageHLA-IM reaches AUROC of 81.4\% and AUPRC of 73.8\% on held-out immunogenicity data, improving over the strongest immunogenicity baselines ImmuneApp (70.7% / 63.6%) and BigMHC-IM (60.1% / 51.5%). With substantially fewer trainable parameters than prior models, TriStageHLA enables efficient, calibrated epitope prioritization across binding, presentation, and immunogenicity. Code and pretrained weights are available at https://github.com/SkywalkerLuke/TriStageHLA.
Related
- Evolutionarily Conserved Plasmodium falciparum HSP70-3 Enhances the Immunogenicity of the Established Malaria Vaccine Antigen Circumsporozoite Protein through Cross-Presentation
- 777 High-throughput discovery of rare, functionally potent antigen-reactive TCRs via polyclonal library screening
- Closing the phenotyping gap: non-invasive electromagnetic induction enables high-throughput screening of crop root function in the field
- An L-threonine-inducible biosensor derived from the tryptophan-repressed trp promoter of Escherichia coli for high-throughput screening of L-threonine overproducers
- High-throughput platforms for genetic perturbation screening using CRISPR/Cas9 in human iPSC-derived macrophages for drug discovery
BibTeX
@misc{t2026unveiling,
title = {{Unveiling immune targets via TriStageHLA: high-throughput screening of peptide presentation, peptide-HLA binding, and immunogenicity}},
author = {T. LU and Z. Wang and Y. Wang and Z. Li and T. Chen and K. Cho and Z. Yao and X. Li and K.-C. Wong},
howpublished = {bioRxiv},
year = {2026},
month = oct,
doi = {10.64898/2026.10.29.762831},
url = {https://www.biorxiv.org/content/10.64898/2026.10.29.762831v1},
note = {CC BY},
}
RIS
TY - UNPB
TI - Unveiling immune targets via TriStageHLA: high-throughput screening of peptide presentation, peptide-HLA binding, and immunogenicity
AU - T. LU
AU - Z. Wang
AU - Y. Wang
AU - Z. Li
AU - T. Chen
AU - K. Cho
AU - Z. Yao
AU - X. Li
AU - K.-C. Wong
PY - 2026
DA - 2026/10/29
PB - bioRxiv
DO - 10.64898/2026.10.29.762831
UR - https://www.biorxiv.org/content/10.64898/2026.10.29.762831v1
AB - High-throughput screening of HLA class I epitopes underpins neoantigen vaccine design, T cell therapies, and pathogen immune monitoring, yet proteome-scale prioritization remains challenging under severe class imbalance. We present TriStageHLA, a compact three-stage framework that decomposes epitope discovery into peptide-only screening (PRE), allele-specific binding/presentation prediction (BIND), and immunogenicity prediction (IM). PRE performs early pruning via a LoRA-adapted ESM-2 classifier; BIND integrates quantitative binding affinity and eluted ligand data with a dual-branch architecture combining global cross-attention and local motif extraction; IM is fine-tuned on curated human T-cell assay data with a decoupled LoRA head. Trained on IEDB-derived datasets with per-allele balancing and allele-discordant supervision, TriStageHLA-BIND achieves AUROC of 96.2% and AUPRC of 95.2% on combined BA/EL evaluation, while TriStageHLA-IM reaches AUROC of 81.4\% and AUPRC of 73.8\% on held-out immunogenicity data, improving over the strongest immunogenicity baselines ImmuneApp (70.7% / 63.6%) and BigMHC-IM (60.1% / 51.5%). With substantially fewer trainable parameters than prior models, TriStageHLA enables efficient, calibrated epitope prioritization across binding, presentation, and immunogenicity. Code and pretrained weights are available at https://github.com/SkywalkerLuke/TriStageHLA.
ER -