Teleport-Stabilized Quantum-Walk Ranking in Near-Tie Neoantigen Regimes
GRIGORIADIS, I.; Emmanouilides, C.
Show abstract
Personalized neoantigen vaccination is a patient-specific decision problem: given a tumors molecular signature--somatic mutations, clonality, RNA expression, and antigen-processing context--we must choose a small, manufacturable peptide set that stays therapeutically relevant under uncertainty. In late-stage pipelines, candidates often collapse into near-ties: binding/presentation estimates, immunogenicity surrogates, and structure-based refinement compress many peptides into narrow score bands, making the final top-K fragile to small shifts in calibration, scaling, sampling, or docking protocols. Similar instability arises in peptide-target discovery when multiple hypotheses remain comparably supported. We introduce a transport-stabilized ranking layer that prioritizes redundancy structure over marginal score differences. Peptides (and structural microstates) become nodes in a patient-conditioned evidence graph; edges encode evidence overlap (motifs/HLA restrictions, processing features, target neighborhoods, pocket/contact fingerprints). We apply symmetry-aware quotient reduction of a normalized graph operator, collapsing near-symmetric neighborhoods into basin units while preserving effective shortlist couplings. Discriminative basin fingerprints are then extracted using coherent quantum-walk transport, |{psi}(t)[>] = e[-]iHt|{psi}(0)[>], with visitation P(v,t) = |[<]v|{psi}(t)[>]|2. Because coherent dynamics are oscillatory and horizon-dependent, we introduce a teleport-consensus channel that mixes unitary transport with restart to yield a stationary marginal suitable for stable ranking,{rho} t+1 = (1 [-] )U{rho}tU{dagger} + {Sigma}jvj|j[>][<]j|, and{pi} i = Tr({Pi}i{rho}). Information-theoretic polygraphs--entropy, dispersion, and consensus traces--quantify stabilization and provide an interpretable tie-breaking audit trail. We demonstrate consistent stabilization across colorectal-cancer contexts spanning peptide-target mechanistic triage, microstate symmetry auditing, multimodal evidence fusion, docking-ensemble geometrization, and patient-specific neoantigen shortlist construction.
Matching journals
The top 6 journals account for 50% of the predicted probability mass.
Similar papers in this journal
- Jaxley: Differentiable simulation enables large-scale training of detailed biophysical models of neural dynamics 95%
- Statistical inference with a manifold-constrained RNA velocity model uncovers cell cycle speed modulations 95%
- Embryo mechanics cartography: inference of 3D force atlases from fluorescence microscopy 94%
Similar papers in this journal
- Distributed information encoding and decoding using self-organized spatial patterns 93%
- Bi-level Graph Learning Unveils Prognosis-Relevant Tumor Microenvironment Patterns in Breast Multiplexed Digital Pathology 93%
- Essential Regression - a generalizable framework for inferring causal latent factors from multi-omic human datasets 93%
"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.