Researchers at Keio University have overcome a significant hurdle in RNA design by implementing a factorization machine with a quadratic optimization annealing (FMQA) framework, providing a more efficient alternative to traditional methods. A team led by Project Lecturer Shuta Kikuchi and Professor Hide Tanaka tackled the challenge of reverse folding and identified RNA sequences that predictably fold into the desired shape. This process is complicated by the exponential increase in possibilities for even short sequences. Their study, published in Scientific Reports on May 3, 2026, revealed that the way RNA sequences are converted into binary format has a significant impact on the success of AI-driven design. “We investigated a new application of FMQA in biomolecular design, where its potential remains relatively untapped,” says Dr. Kikuchi, emphasizing that this work explores an unexplored aspect of biomolecular engineering.
New computational frameworks reduce the time and cost of designing functional RNA molecules, providing potential advances in therapeutics. Traditional methods struggle with exponential growth in probability even for relatively short sequences and require extensive computational resources and experimental validation. Kikuchi discovered that how the RNA sequence is converted into a binary representation has a significant impact on optimization performance. Specifically, one-hot and domain-wall encoding consistently performed better than binary and uninary techniques, resulting in sequences with improved thermodynamic stability. The researchers found that assigning guanine (G) and cytosine to preferred states within the domain wall code promoted the formation of G,C base pairs in the stem region. Across benchmarks, FMQA identified high-quality RNA designs with fewer evaluations than competing methods such as random search, genetic algorithms, and Bayesian optimization. Dr. Kikuchi suggests potential applications for biosensors, genome editing tools, and aptamers, and the research team predicts that this highly evaluation-efficient approach will accelerate discoveries across biotechnology and medicine, potentially leading to the use of quantum annealing machines in the future.
We investigated new applications of FMQA in biomolecule design, whose potential remains relatively untapped. Because RNA, DNA, and protein sequences are categorical in nature, it is unclear how converting them to a binary representation will affect optimization performance. In this study, we investigated RNA inversion and the effects of different encoding and assignment choices within FMQA.
The pursuit of rationally designed RNA molecules essential for therapeutics is increasing our reliance on artificial intelligence, but how biological sequences are translated into language that AI can understand, or encoding strategies, remains a surprisingly understudied variable. Quality was assessed using normalized ensemble defects (NED), a measure of structural agreement. The results revealed a clear hierarchy of performance. Dr. Kikuchi explained that one-hot encoding and domain-wall encoding always perform better than their counterparts, producing sequences with lower NED values and higher success rates. This efficiency, coupled with the possibility of implementation in quantum annealing machines, suggests a path to accelerating the design of biomolecules and reducing the experimental burden on researchers. “They have the versatility to be applied to discrete design problems, such as materials and molecular design, where each evaluation is costly,” adds Professor Tanaka.
They are versatile enough to be applied to discrete design problems, such as materials and molecular design, where each evaluation is costly.
Researchers at Keio University have refined the metrics used to assess the quality of artificially designed RNA sequences, moving beyond simple accuracy to a more nuanced understanding of structural fidelity. Focusing on structural matching is critical for therapeutic applications, as even small deviations can significantly change the functionality of a molecule. Their research shows that simply achieving sequence is not enough. Sequence quality, as defined by NED, is of paramount importance. Kikuchi also revealed the surprising impact of nucleotide assignment within the coding scheme, emphasizing that data coding is not just a preliminary step, but rather “a design variable that can fundamentally shape the outcome of the optimization.” The team’s findings suggest that a more holistic approach to RNA design that considers both optimization algorithms and encoding strategies is essential to accelerate the development of functional biomolecules, from biosensors to gene editing tools.
Potential applications include biosensors, genome editing tools, aptamers, ribozymes, and riboswitches.
This success is not simply due to the optimization algorithm itself. Through their investigation of coding strategies, the researchers discovered previously “unexplored aspects of biomolecular engineering.” When guanine and cytosine were assigned to these favorable states, the resulting RNA sequences exhibited better thermodynamic stability, an important factor for therapeutic applications. Looking to the future, researchers envision a future where FMQA is integrated with quantum computing technology. “FMQA formulates the learned surrogate model as a quadratic optimization problem, so it can be implemented with a quantum annealing machine,” says Dr. Kikuchi, pointing to the potential for “advancing ‘Quantum for Biology.’” This marriage of quantum-inspired computing and biomolecular engineering has the potential to accelerate the design of functional biomolecules, biosensors, genome editing tools, and aptamers, ultimately reducing the experimental burden on biotechnology and medicine.
FMQA formulates the learned surrogate model as a quadratic optimization problem, so it can be implemented with a quantum annealing machine.
