American researchers from IBM and Moderna have used quantum modeling to predict the complex secondary structure of a 60-nucleotide mRNA sequence, the longest ever simulated on a quantum computer.\n\nmRNA — matrix information ribonucleic acid — a molecule, that carries genetic information from DNA to ribosomes. It controls the synthesis of protein in cells and is used to create vaccines that elicit a specific immune response.\n\nIt is widely believed, that all the information, necessary for a protein, to acquire a three-dimensional conformation, is contained in its amino acid sequence. And although mRNA consists of a single chain of amino acids, it has a secondary protein structure consisting of from a series of folds that provide a specific three-dimensional shape of this molecule.\n\nThe number of possible folding variants increases exponentially with each added nucleotide. This makes it difficult to predict the shape of the molecule RNA on a larger scale. An experiment conducted by IBM and Moderna back in 2024 demonstrated how quantum computing can be used as an additional tool for forecasting.\n\nTraditionally, such predictions have been based on binary classical computers and artificial intelligence models such as Google DeepMind’s AlphaFold. The new study notes, that the algorithms of these classical architectures can process mRNA sequences consisting of hundreds or thousands of nucleotides, but exclude more complex elements such as \"pseudo-nodes\".\n\nPseudo-junctions are complex bends and shapes in the secondary structure of a molecule, that are capable of more complex internal interactions than ordinary folds. Due to their exclusion, the potential accuracy of any protein folding prediction model is fundamentally limited.\n\nUnderstanding and predicting the smallest details of the protein structure of a molecule mRNA is key to developing more reliable predictions and more effective mRNA-based vaccines. The researchers hope to overcome the limitations, inherent in the most powerful supercomputers and AI models.\n\nScientists have conducted a large number of experiments, using quantum modeling algorithms based on qubits, to simulate the structure of molecules mRNA. Initially, they used only 80 qubits out of a possible 156 on the quantum processor R2 Heron. The researchers used variational quantum algorithm, based on conditional value at risk — a quantum optimization algorithm modeled after certain methods, used for analyzing complex interactions to predict the secondary structure of a protein from a 60-nucleotide mRNA sequence.\n\nBefore that, the best result for the quantum simulation model was a 42-nucleotide sequence. The researchers scaled up the experiment by applying modern error correction methods to eliminate noise generated by quantum fluctuations.\n\nThe study demonstrated the possibility of effective use of quantum computing to predict the structure of mRNA with a length of 60 nucleotides. Scientists also conducted a preliminary study that demonstrated the possibility of using up to 354 qubits for the same algorithms in the absence of noise.\n\nAccording to the authors of the study, an increase in the number of qubits used to run the algorithm while scaling algorithms for additional subroutines should significantly improve modeling accuracy and the ability to predict longer sequences. However these methods require the development of advanced technologies to embed these problem-oriented schemes in existing quantum equipment.\n\nThe results of the study are published on the preprint server arXiv.\n\nSource: LiveScience

A quantum computer has defeated AI: IBM and Moderna simulate the longest mRNA molecule

American researchers from IBM and Moderna have used quantum computing to simulate the secondary structure of a 60-nucleotide mRNA sequence, the longest ever achieved on a quantum computer. The simulation leveraged a variational quantum algorithm with error correction on IBM's R2 Heron processor, which has 156 qubits. This marks a significant advancement over previous quantum simulations of 42-nucleotide sequences and highlights the potential of quantum computing to model complex molecular folding, particularly pseudo-junctions that classical AI models like AlphaFold cannot fully capture. The study, published on arXiv, suggests that scaling up to 354 qubits in noise-free conditions could further improve accuracy and enable prediction of longer mRNA sequences, though practical implementation requires advances in quantum hardware and error mitigation.