TAMAVAQ/Q‐TAMAVAQ: Fubini–study CTQW graph transport and constrained margin‐based ranking for neoantigen candidate prioritization in lower‐grade glioma and glioblastoma
Ioannis GrigoriadisAbstract
Background
Personalized neoantigen vaccine design for glioma requires the integration of mutation‐derived peptide evidence, molecular properties, HLA coverage and finite manufacturing capacity. TAMAVAQ/Q‐TAMAVAQ combines quantum‐geometric descriptor representation with graph transport and auditable peptide‐panel assembly.
Methods
A 7262‐row lower‐grade glioma/glioblastoma source ledger was reconciled to 4128 unique variant–peptide–HLA candidates. A separate Seq2Neo benchmark contained 1537 peptide–HLA records, including 65 positives defined by a fixed predictor‐derived composite endpoint. Normalized molecular descriptors were encoded by squared‐fidelity Fubini–Study distances and a weighted graph Laplacian. Continuous‐time quantum‐walk (CTQW) transport and classical heat diffusion were evaluated on identical graphs across 60 combinations of neighbourhood size, bandwidth and propagation time. Non‐compensatory biological, structural, geometric, manufacturing, provenance and validation gates were coupled to deterministic margin‐based HLA‐aware panel selection.
Results
The integrated ranking achieved average precision (AP) 0.552, AUROC 0.962 and NDCG@100 0.693 on the internal Seq2Neo endpoint, recovering 17, 33 and 45 positives at ranks 25, 50 and 100, respectively. At the matched reference setting, CTQW and heat diffusion yielded AP 0.481 and 0.471, NDCG@100 0.637 and 0.632, and AUROC 0.951 and 0.956, respectively. CTQW yielded higher AP in 15 of 60 parameter configurations. A 14‐candidate mechanistic register demonstrated deterministic four‐peptide selection, admissible reserves and provenance‐preserving replay. Exploratory FCS analysis supplied a separate descriptive immune‐phenotyping layer.
Conclusions
The integrated workflow links molecular dossier geometry, complementary graph–transport operators and explicit selection constraints to a traceable, HLA‐aware shortlist of glioma neoantigen candidates.