Rapid, accurate in silico neoantigen identification
Neoantigens—mutated peptides arising from tumor-specific genomic alterations—are key targets for personalized cancer vaccines and T-cell therapies. Identifying immunogenic neoantigens from the vast pool of somatic mutations remains a bottleneck: traditional binding-affinity filters yield high false-positive rates and fail to prioritize candidates that actually elicit T-cell responses. Our AI Neoantigen Prediction Service addresses this challenge with a deep neural network framework that integrates peptide-MHC binding affinity, antigen-processing signals, and immunogenicity features to rank the most promising neoantigens.
The service begins with tumor and normal sequencing data (WES/WGS/RNA-seq) to call somatic variants—including SNVs, indels, and gene fusions—and perform HLA typing. An ensemble of AI models trained on large-scale peptide-MHC and mass spectrometry datasets then predicts MHC class I (and where applicable, class II) presentation and immunogenic potential. The output is a prioritized list of neoantigen candidates with supporting evidence, enabling downstream validation and rapid translation into clinical candidate selection.
What we deliver
Comprehensive neoantigen candidate list
A curated list of candidate peptides from all identified somatic mutations, with predictions for MHC class I binding and immunogenicity, ranked by confidence score.
High-confidence prioritization
Candidates are filtered and prioritized using a multi-layer AI model that accounts for HLA binding affinity, proteasomal cleavage, TAP transport, and T-cell recognition potential.
Validation recommendations
Each report includes suggested experimental validation assays (e.g., tetramer binding, ELISpot) tailored to the top candidates, accelerating downstream verification.
How the service works
Service scope
| Parameter | Typical project scope |
|---|---|
| Input data | Scoped per project after consultation |
| Mutation analysis | Somatic SNV, indel, and fusion detection; variant annotation with population frequency filtering is performed to identify tumor-specific mutations. |
| HLA typing | HLA class I (A, B, C) and class II (DRB1, DQB1, DPB1) typing is performed from sequencing data, with allele confidence scoring. |
| Peptide-MHC binding prediction | AI-based prediction is conducted for 8–11 mer peptides (class I) and 15–20 mer (class II), covering multiple HLA alleles per patient. |
| Immunogenicity score | A composite score integrating binding affinity, processing signals, and T-cell recognition propensity is provided for each candidate. |
Why choose our AI-driven approach
Deep learning ensemble
Our prediction engine combines multiple neural network architectures trained on large peptide-MHC binding and mass spectrometry elution datasets, capturing non-linear features that linear models miss.
Multi-feature integration
Beyond simple binding affinity, the model incorporates proteasomal cleavage, TAP transport efficiency, and immunogenicity metrics to reduce false positives and prioritize truly actionable neoantigens.
Rapid throughput
Fully automated pipeline from sequencing data to prioritized candidate list in weeks, not months—enabling faster decision-making for personalized vaccine or TCR therapy development.
Expert scientific support
Each project is managed by a computational immunologist who interprets results, discusses candidate selection, and provides validation strategy recommendations tailored to your therapeutic program.
Supporting literature
Read the method
Computational Methods for Cancer Neoantigen Prediction – Springer Nature Protocol
View articleReady to accelerate your neoantigen discovery?
Submit your patient samples and receive a prioritized list of AI-predicted neoantigen candidates with full supporting data and validation recommendations. Contact our team to discuss your project scope and timeline.