Q1Fold: A Qubit-Efficient Hybrid Quantum-Classical Convolutional Neural Network for RNA secondary Structure Prediction

Zeyu Li, Kang Han, Wei Xiang

Keywords: Quantum Machine Learning, Variational Quantum Circuit, RNA 2D structure prediction

Abstract: RNA 2D structure prediction remains a critical challenge in computational biology, with existing thermodynamic and deep learning approaches facing limitations in modeling complex interactions and data requirements. We introduce Q1Fold, a hybrid quantum-classical convolutional network for RNA secondary structure prediction. The model integrates a compact variational quantum circuit with a classical 2D ResNet architecture, where the quantum circuit generates expressive features from local sequence windows using minimal qubits. This design avoids barren plateaus and is compatible with current Noisy Intermediate-Scale Quantum Computers. Despite using significantly fewer parameters, Q1Fold achieves competitive performance on standard benchmarks compared to state-of-the-art methods. The extracted quantum features also demonstrate superior capability in representing local structural motifs such as hairpins. Our work establishes a practical route toward quantum-enhanced computational RNA biology.

Primary Area: applications to physical sciences (physics, chemistry, biology, etc.)

Submission Number: 10016