A Hybrid Quantum-AI Framework for Protein Structure Prediction on NISQ Devices

TL;DR

This work tackles protein structure prediction on near-term quantum hardware by marrying first-principles VQE-derived conformations with data-driven structural priors. A five-stage framework generates quantum candidates on a 127-qubit IBM processor, enriches them with NetSurfP-3.0 priors, and performs energy fusion via $E_{\mathrm{fuse}}(c)=\alpha\widetilde{E}_q(c)+\beta\widetilde{D}_{\mathrm{ss}}(c)+\gamma\widetilde{D}_{\angle}(c)$ to re-rank conformations. The fused energy framework sharpens coarse quantum landscapes with high-resolution learned potentials, yielding statistically significant RMSD improvements over AlphaFold3, ColabFold, and purely quantum baselines (mean RMSD $=4.89$ Å, $p<0.001$). Demonstrating on 75 protein fragments, the approach provides a practical path toward hybrid quantum-classical computational biology, highlighting a general strategy to integrate physics-based quantum predictions with neural priors for robust biomolecular modeling.

Abstract

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 consistent improvements over AlphaFold3, ColabFold, and quantum-only predictions, achieving a mean RMSD of 4.9 Å with statistical significance (p < 0.001). 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.

Figures (6)

• Figure 1: Overview of the five-stage framework for quantum protein structure prediction. The workflow proceeds from sequence input through encoding, quantum energy minimization, refinement, and evaluation. Stages (ii) and (iv) enable hybrid operation by incorporating biologically informed priors derived from neural models. This systematic design provides a reproducible engineering structure that unifies pure quantum prediction and hybrid quantum and classical strategies.
• Figure 2: Overview of the proposed hybrid framework. Candidate conformations are generated using VQE on quantum hardware, augmented with structural features from deep learning models, and re-ranked via fused energy terms to produce the final prediction.
• Figure 3: Comparison of energy landscapes. (a) Quantum-derived energy landscape shows multiple coarse-grained basins. (b) Deep-learning priors contribute fine-grained local refinements in valley regions. (c) The fused energy landscape sharpens valley resolution by combining quantum basins with DL-detailed gradients, enabling more accurate identification of native-like structures.
• Figure 4: Comprehensive comparison of AF3, ColabFold, Quantum, and Hybrid methods in terms of RMSD. Results demonstrate that the Hybrid framework consistently outperforms both classical AI-based approaches (AF3, ColabFold) and pure quantum predictions, achieving lower errors across distributions, cumulative statistics, per-sample improvements, and overall win counts.
• Figure 6: Case study of score composition in four representative protein fragments. Each stacked bar plot shows contributions from quantum energy ($E_q$, blue), secondary-structure distribution difference ($D_{ss}$, green), and torsional angle difference ($D_{\phi\psi}$, orange). These examples highlight that all three terms are necessary and complementary: while quantum energy provides a baseline signal, structural priors refine and stabilize the ranking, ensuring that the hybrid method consistently selects the most accurate conformation.