Neoantigen prediction as the navigation layer for precision immunotherapy

The “navigation layer” for precision immunotherapy

In individualized oncology immunotherapy, neoantigen prediction is a foundational step that influences whether downstream vaccine and cell-therapy designs can engage the right T cell responses. Runsheng Cell describes a multi-omics + AI stack spanning mutation calling, expression context, antigen processing presentation modeling, and T cell immunogenicity assessment.

Program highlights (as described in company materials)

1. Dual-algorithm engine: couples deep learning with biophysical models to balance annotation depth and HLA coverage (materials cite ~30% higher positive-antigen hit rate—requires independent validation).
2. Noise control: RNA-seq thresholds, monoallelic expression correction, and heterogeneity indices to reduce false positives (materials cite AUC ~0.93—cohort-dependent).
3. Turnaround: distributed in-memory compute targeting ~10 hours end-to-end for prediction cycles described by the team.
4. Broader variant classes: fusion genes, splicing events, intron retention, and TCR-informed discovery pathways beyond SNVs/indels—scope as claimed in filings.
5. Continuous iteration: model updates with new public data and clinical feedback; emphasis on rare mutations and extended HLA coverage.

Patents (summary)

• ZL202410515132.1 — NGS-based neoantigen identification method/system
• ZL202411175913.7 — Atypical neoantigen screening

Patent titles are summarized; claims define legal scope.

Informational article; performance metrics should be verified against peer-reviewed studies and regulatory communications.