Abstract Recent advances in deep learning have enabled highly accurate protein structure prediction, yet predicted structures frequently exhibit localized geometric pathologies such as steric clashes, distortions, and poorly formed regions that limit downstream use. These errors are typically rare but high-impact, suggesting that refinement is a tail-dominated problem rather than a uniform optimization task. We propose Q-Refine-AF, a hybrid refinement framework that formulates post-prediction structure correction as an operator learning problem acting on local structural patches. Patch selection is driven by a tail-risk objective inspired by Conditional Value-at-Risk (CVaR), ensuring that refinement effort is concentrated on the most problematic regions while leaving well-formed parts of the structure untouched. The refinement operator is implemented as a parameterized quantum circuit embedded within a classical pipeline and trained to produce structured geometric updates rather than scalar quality scores. Through controlled synthetic experiments, we demonstrate that the proposed framework exhibits two distinct operating regimes: a conservative regime characterized by near-identity updates that preserve structural stability, and an aggressive regime capable of resolving severe synthetic steric conflicts through large corrective displacements. While the latter regime substantially reduces clash counts in highly corrupted structures, the resulting updates are not yet physically constrained and should be interpreted as structural repair rather than biophysically realistic refinement. These results validate the core hypothesis that tail-risk–aware operator learning can selectively target dominant local failure modes in structure prediction pipelines. Although no quantum computational advantage is claimed and physical constraints remain future work, QRefine-AF establishes a foundation for integrating learned, selective refinement operators with existing protein structure prediction and relaxation workflows.
Abstract Recent advances in deep learning have enabled highly accurate protein structure prediction, yet predicted structures frequently exhibit localized geometric pathologies such as steric clashes, distortions, and poorly formed regions that limit downstream use. These errors are typically rare but high-impact, suggesting that refinement is a tail-dominated problem rather than a uniform optimization task. We propose Q-Refine-AF, a hybrid refinement framework that formulates post-prediction structure correction as an operator learning problem acting on local structural patches. Patch selection is driven by a tail-risk objective inspired by Conditional Value-at-Risk (CVaR), ensuring that refinement effort is concentrated on the most problematic regions while leaving well-formed parts of the structure untouched. The refinement operator is implemented as a parameterized quantum circuit embedded within a classical pipeline and trained to produce structured geometric updates rather than scalar quality scores. Through controlled synthetic experiments, we demonstrate that the proposed framework exhibits two distinct operating regimes: a conservative regime characterized by near-identity updates that preserve structural stability, and an aggressive regime capable of resolving severe synthetic steric conflicts through large corrective displacements. While the latter regime substantially reduces clash counts in highly corrupted structures, the resulting updates are not yet physically constrained and should be interpreted as structural repair rather than biophysically realistic refinement. These results validate the core hypothesis that tail-risk–aware operator learning can selectively target dominant local failure modes in structure prediction pipelines. Although no quantum computational advantage is claimed and physical constraints remain future work, QRefine-AF establishes a foundation for integrating learned, selective refinement operators with existing protein structure prediction and relaxation workflows.
Abstract Protein structure prediction and decoy ranking remain central challenges in computational biophysics. Classical scoring functions often struggle todiscriminate near-native conformations from large populations of plausible decoys,particularly in the critical low-energy tail of the conformational distribution. Weintroduce Q-CVaR-Fold, a hybrid quantum–classical architecture that integratesa geometric graph neural network encoder with a small parameterized quantumcircuit acting as a nonlinear scoring head. To focus optimization on near-nativeconformations, we combine contrastive ranking with Conditional Value-at-Risk(CVaR) tail reweighting, yielding a risk-sensitive training objective aligned withstructural evaluation metrics.Despite using only four qubits and shallow entangling layers, Q-CVaR-Fold exhibits stable end-to-end training and avoids barren plateaus. On a decoy-rankingbenchmark, the model achieves a ROC-AUC of 0.984 and perfect top-5 nativeenrichment across all sequences, outperforming classical baselines of comparablesize. The score distributions and monotonic reduction of CVaR loss demonstratethat quantum feature transformations, coupled with tail-focused optimization, provide discriminative power beyond standard MLP heads. To our knowledge, this isthe first demonstration of a quantum-enhanced, risk-sensitive scoring model thatachieves near-perfect recovery of native structures in decoy-ranking tasks.Q-CVaR-Fold highlights the potential of hybrid quantum models for energylandscape modeling, fragment selection, and structural refinement, and offers apromising foundation for next-generation quantum–geometric methods in computational structural biology.
Protein structure prediction remains central to computational biology and drug discovery. While deep learning models such as AlphaFold have achieved remarkable accuracy for known fold families, they struggle with novel folds, intrinsically disordered regions, and conformational dynamics, and they do not explicitly optimize physics-based energy functions. We present a quantum-assisted approach to protein conformation optimization that uses parameterized quantum circuits within a variational framework to explore the energy landscape of small proteins and fragments from first-principles or coarse-grained physics. Our method encodes backbone torsion angles (ϕ, ψ) and selected side-chain degrees of freedom into qubit registers and employs a variational quantum eigensolver (VQE)-inspired optimization to minimize an approximate Hamiltonian incorporating bonded interactions, sterics, hydrogen bonding, and hydrophobic contacts. We introduce problem-inspired ansätze that respect chain connectivity and secondary-structure motifs (helices, beta-sheets) and propose hierarchical circuit growth strategies for scaling from peptides to larger domains. Through classical simulations of small peptides and secondary-structure motifs, we demonstrate that quantum variational optimization can identify low-energy conformations consistent with experimentally determined structures. We provide explicit resource estimates for NISQ and early fault-tolerant quantum computers and discuss hybrid classical-quantum strategies for refining quantum solutions with classical molecular dynamics and machinelearning priors. While full protein-folding quantum advantage remains distant, we identify concrete subproblems where nearterm quantum devices offer complementary physics-based insight beyond data-driven approaches, opening pathways toward practical quantum-classical protein structure and drug discovery pipelines.
Accurate prediction of mRNA secondary structure is critical for understanding gene expression, translation efficiency, and advancing mRNA-based therapeutics. However, the combinatorial complexity of possible foldings, especially in long sequences, poses significant computational challenges for classical algorithms. In this work, we propose a scalable, quantum-centric optimization framework that integrates quantum sampling with classical post-processing to tackle this problem. Building on a Quadratic Unconstrained Binary Optimization (QUBO) formulation of the mRNA folding task, we develop two complementary workflows: a Conditional Value at Risk (CVaR)-based variational quantum algorithm enhanced with gauge transformations and local search, and an Instantaneous Quantum Polynomial (IQP) circuit-based scheme where training is done classically and sampling is delegated to quantum hardware. We demonstrate the effectiveness of these approaches using IBM quantum processors, solving problem instances with up to 156 qubits and circuits containing up to 950 nonlocal gates, corresponding to mRNA sequences of up to 60 nucleotides. Additionally, we validate scalability of the CVaR algorithm on a tensor network simulator, reaching up to 354 qubits in noiseless settings. These results demonstrate the growing practical capabilities of hybrid quantum-classical methods for tackling large-scale biological optimization problems.
Abstract Near-term quantum devices remain constrained by limited qubit counts, noise, and shallow circuit depth, restricting their direct application to peptide folding. This work addresses this limitation through a hybrid quantum–classical framework that separates global contact-space exploration from local configuration-space refinement. The pipeline was evaluated on the p53 transactivation peptide. The peptide was encoded using discrete torsional states and Miyazawa–Jernigan contact potentials. Global conformational search was performed using simulated annealing, followed by fragment-level CVaR-based variational quantum eigensolver refinement. Overlapping fragments were stitched and reconstructed into three-dimensional backbone coordinates for structural comparison. Three independent simulated annealing restarts produced negative-energy conformers. The workflow identified a best global conformer at E = −3.70 MJ using simulated annealing. Fragment-level CVaR-VQE refinement generated 30 candidate conformers across three overlapping fragments, the best assembled conformer reaching E = −2.58 MJ after DP stitching and post-stitch refinement. In this trial, the classical SA solution outperformed the VQE-assembled structure and was selected as the final output. The final 65-atom model was successfully reconstructed, but full-backbone alignment against 1YCR yielded an RMSD of 4.65 Å, improving to 2.90 Å after trimmed alignment. Our work demonstrates a practical quantum-enabled peptide folding workflow for short, flexible peptides, but the present structural accuracy is insufficient for claims of native fold recovery. Its main value is as an auditable framework for testing NISQ-compatible folding strategies.
Accurate prediction of binding affinities and efficient lead optimization remain central challenges in computational drug discovery, particularly for complex biological targets characterized by rugged energy landscapes, conformational heterogeneity, and high-dimensional interaction spaces. Classical computational methods—such as molecular docking, force-field–based simulations, and deep learning models—have made substantial progress, yet they often struggle to capture quantum-mechanical effects and long-range correlations critical to molecular recognition. This study explores a quantum-enhanced framework that integrates Variational Quantum Algorithms (VQAs), quantum-inspired Hamiltonian modeling, and hybrid quantum–classical optimization workflows to improve the fidelity and scalability of drug discovery pipelines. By encoding molecular interactions into parameterized quantum circuits, the framework leverages quantum superposition and entanglement to explore complex chemical space more efficiently, enabling refined estimation of binding energies and accelerated identification of high-quality leads. Preliminary simulations indicate that VQA-based estimators outperform classical baselines for challenging protein–ligand systems, reducing prediction error while maintaining computational tractability. These findings highlight the emerging potential of quantum technologies to enable more accurate, data-efficient, and mechanistically grounded drug discovery for next-generation therapeutics.
Variational quantum algorithms provide a direct, physics-based approach to protein structure prediction, but their accuracy is limited by the coarse resolution of the energy landscapes generated on current noisy devices. We propose a hybrid framework that combines quantum computation with deep learning, formulating structure prediction as a problem of energy fusion. Candidate conformations are obtained through the Variational Quantum Eigensolver (VQE) executed on IBM’s 127 qubit superconducting processor, which defines a global yet low resolution quantum energy surface. To refine these basins, secondary structure probabilities and dihedral angle distributions predicted by the NSP3 neural network are incorporated as statistical potentials. These additional terms sharpen the valleys of the quantum landscape, resulting in a fused energy function that enhances effective resolution and better distinguishes native like structures. Evaluation on 375 conformations from 75 protein fragments shows that, on this short fragment-only benchmark (10–14 residues without surrounding chain context), the hybrid pipeline achieves a mean RMSD of 4.9 Å with statistical significance ( p < 0.001), outperforming both quantum-only re-rankings and fragment-level runs of AlphaFold3 and ColabFold. We caution that this conclusion is restricted to the fragment-scale regime accessible to current quantum hardware and does not extend to the full-chain regime in which AlphaFold3 and ColabFold are usually evaluated. The findings demonstrate that energy fusion offers a systematic method for combining data driven models with quantum algorithms, improving the practical applicability of near term quantum computing to molecular and structural biology.