Fitness landscapes and thermodynamic approaches to development of nucleic acids enzymes: from classical methods to AI integration.

Nucleic acids (NA), namely DNA and RNA, dynamically fold and unfold to perform their functions in cells. Functional NAs include NA enzymes, such as ribozymes and DNAzymes. Their folding and target binding are governed by interactions between nucleobases, including base pairings, which follow thermodynamic principles. To elucidate biological mechanisms and enable diverse technical applications, it is essential to clarify the relationship between the primary sequence and the catalytic activity of NA enzymes. Unlike methods for predicting the stability of NA duplexes, which have been widely used for over half a century, predictive approaches for the catalytic activity of NA enzymes remain limited due to the low throughput of activity assays. However, recent advances in genome analysis and computational data science have significantly improved our understanding of the sequence-function relationship in NA enzymes. This article reviews the contributions of data-driven chemistry to understanding the reaction mechanisms of NA enzymes at the nucleotide level and predicting novel NA enzymes with catalytic activity from sequence information. Furthermore, we discuss potential databases for predicting NA enzyme activity under various solution conditions and their integration with artificial intelligence for future applications.