An Overview and Benchmark Evaluation of RNA Representation Learning Methods

Representation learning methods are increasingly essential in biological sequence analysis, efficiently extracting features from extensive sequence datasets. While prior surveys have concentrated on proteins, benchmarks for RNA sequences are notably absent. In this study, we review specific RNA representation learning methods and establish a benchmark focused on downstream tasks including RNA-protein binding prediction and microRNA subcellular localization prediction. Our comparative analysis of these representation learning methods provides new insights into their efficacy in nucleic acid molecular function prediction.