An unexpectedly effective Monte Carlo technique for the RNA inverse folding problem

Solving the RNA inverse folding problem, also known as the RNA design problem, is critical to advance several scientific fields like bioengineering, yet existing approaches have had limited success. The problem has several features that resist traditional computational techniques, such as its exponential complexity and the chaotic behavior of its cost function. Although some state-of-the-art AI approaches have reported promising results, all existing computational methods substantially underperform expert human designers. I combine a different technique, Nested Monte Carlo Search (NMCS), with domain-specific knowledge to create an algorithm that outperforms all prior published methods by wide margins and solves 95 of the 100 puzzles listed in a recently proposed RNA solving difficulty benchmark.