Towards secondary structure prediction of longer mRNA sequences using a quantum-centric optimization scheme

Elucidating mRNA secondary structure is crucial for advancing mRNA synthesis, translation regulation, and therapeutic innovation, yet it faces formidable computational challenges. This study introduces a quantum-centric approach that significantly enhances the computational feasibility of predicting mRNA secondary structures. By effectively balancing tasks between quantum processors and classical computing nodes, our methods utilize a quantum sampling-based paradigm to arrive at the optimal solution. Here, we advance our previous work by integrating a variational CVaR-based quantum algorithm with a novel noise mitigation and problem reduction strategy. We further validate the scalability of this approach on a tensor-network simulator for problems with up to 358 qubits, demonstrating its robustness without noise. Additionally, we utilize Instantaneous Quantum Polynomial circuits, which facilitate classical computations of expectations to guide the training of variational circuits and also aid in error mitigation. These circuits support computationally intensive sampling tasks on quantum hardware. Our scheme is complemented by a lightweight classical local search that refines these samples, thereby increasing the likelihood of identifying optimal solutions. We present results that validate both techniques on IBM quantum processors by conducting hardware experiments that require up to 156 qubits and 950 nonlocal gates, effectively solving mRNA sequences of up to 60 nucleotides. These advancements not only demonstrate the potential of quantum workflows to solve complex computational problems in biology but also pave the way for groundbreaking applications in bioinformatics and therapeutic development.