Review of machine learning methods for RNA secondary structure prediction

Abstract

Secondary structure plays an important role in determining the function of noncoding RNAs. Hence, identifying RNA secondary structures is of great value to research. Computational prediction is a mainstream approach for predicting RNA secondary structure. Unfortunately, even though new methods have been proposed over the past 40 years, the performance of computational prediction methods has stagnated in the last decade. Recently, with the increasing availability of RNA structure data, new methods based on machine learning (ML) technologies, especially deep learning, have alleviated the issue. In this review, we provide a comprehensive overview of RNA secondary structure prediction methods based on ML technologies and a tabularized summary of the most important methods in this field. The current pending challenges in the field of RNA secondary structure prediction and future trends are also discussed.

Quality by Design for enabling RNA platform production processes

“…Similarly, a powerful LNP degradation model, embracing thermodynamic stability, aggregation, and physical degradation, could predict both in-process and in vivo stability. Following the same logic, the available RNA sequence design strategy, underpinned by various data-driven, hybrid, and molecular dynamic modeling tools, could consider the sequence manufacturability in addition to RNA stability and translation [41,[115][116][117]. The secondary structure is indeed known to impact in-process degradation and capping accessibility and could play a potential role in LNP formation [37].…”

Probing RNA structures and functions by solvent accessibility: an overview from experimental and computational perspectives

“…Examples are mfold [ 18 ], the first MFE-based RNA secondary structure prediction tool, the Vienna RNA package [ 19 ], UNAfold [ 20 ] and RNAstructure [ 21 ]. Recent advances in deep learning allow end-to-end prediction of RNA base-pairing structures with improved performance not only in canonical base pairs but also in pseudoknots and non-canonical base pairs associated with tertiary interactions [ 22–25 ].…”

“…Here, we will provide an overview of recent progresses in experimental and computational approaches to RNA SASA as it has been an overlooked area of research. Other recent reviews on studies of RNA secondary or tertiary structures can be found elsewhere [ 14 , 22 , 30 , 31 ].…”

Decisive Roles of Sequence Distributions in the Generalizability ofde novoDeep Learning Models for RNA Secondary Structure Prediction

“…As such, there have been major interests in determining and understanding RNA secondary structures, via both experiment and computation (10)(11)(12). In recent years, with the emergence of sizeable RNA structure databases and the accessibility of powerful machine learning algorithms, data-centric deep-learning-based models, the subject of this study, have been successfully developed for RNA secondary structure prediction (13).…”

“…Varieties of efficient algorithms, especially based on dynamic programming and related techniques (17), have also been introduced along with improved scoring parameters (15,16). However, traditional algorithms have struggled to make significant gains in performance in the recent decade (13).…”