Direct RNA sequencing and signal alignment reveal RNA structure ensembles in a eukaryotic cell

nature methods Article https://doi.org/10.1038/s41592-026-03069-y Direct RNA sequencing and signal alignment reveal RNA structure ensembles in a eukaryotic cell Received: 15 February 2025 Accepted: 19 March 2026 Jiaxu Wang 1,2,5 , Jian Han2,5, Wen Ting Tan2, Anthony Youzhi Cheng2, Jong Ghut Ashley Aw2, Yue Wang 3,4, Guisheng Zeng3, Niranjan Nagarajan 2,4 & Yue Wan 2 Published online: xx xx xxxx Check for updates The extent to which an RNA folds into structure ensembles and how different structures in the ensemble regulate eukaryotic gene expression is not fully understood. Here, we coupled chemical probing with direct RNA sequencing to identify structure modifications along a single RNA molecule (sm-PORE-cupine). We used direct signal alignment in addition to base mapping to increase the percentage of mappable sequences and showed that Bernoulli mixture model clustering can separate structure ensembles accurately. We applied sm-PORE-cupine to identify isoform-specific structure ensembles along the SARS-CoV-2 genome and structure ensembles in the Candida albicans transcriptome. We observed that RNAs are more structurally homogeneous in vitro, at higher temperatures and in the 3′ untranslated regions of C. albicans. Structure ensembles are associated with changes in translation efficiency and decay in C. albicans, and we validated translation changes using reporter assays. sm-PORE-cupine expands the existing toolbox for studying RNA structure and function in diverse transcriptomes. Although there are many examples of functional RNA structures in bacteria and viruses, how RNA structures could regulate gene expression in eukaryotic cells remains an important open question. As RNAs can fold into different conformations in vitro and in vivo, identifying functional RNA structures and how they change and determining which structure in a population ensemble is biologically important enables us to understand fundamental RNA biology and how RNA can be targeted in diseases1,2. To obtain single-molecule RNA structure information, several methods, including DRACO3, DREEM4, DANCE-MaP5 and Da Vinci6, have been developed to couple dimethyl sulfate (DMS) or selective 2′-hydroxyl acylation analyzed by primer extension (SHAPE) chemical probing with high-throughput sequencing to read out RNA modifications along a single read of RNA. These modified patterns are then clustered using different algorithms to identify potential structure clusters that are present inside cells or in solution. Single-molecule RNA structure studies have identified structure ensembles along viral RNAs, such as in HIV4 and SARS-CoV-2 genomes3,7; human RNAs, such as in 7SK5; and plant RNAs, including in COOLAIR6, to better study functional RNA structures that could impact gene regulation. Although useful, most of these strategies require the conversion of RNA into cDNA molecules for analysis. We previously coupled SHAPE chemical probing with nanopore direct RNA sequencing to determine aggregate RNA structure Institute of Medical Genetics and Development, Key Laboratory of Reproductive Genetics (Ministry of Education) and Women’s Hospital, Zhejiang University School of Medicine, Zhejiang, China. 2Genome Institute of Singapore, A*STAR, Singapore, Singapore. 3A*STAR Infectious Diseases Labs (A*STAR ID Labs), Agency for Science, Technology and Research (A*STAR), Singapore, Singapore. 4Department of Biochemistry, Yong Loo Lin School of Medicine, National University of Singapore, Singapore, Singapore. 5These authors contributed equally: Jiaxu Wang, Jian Han. e-mail: ; ; 1 Nature Methods Article information on gene-linked RNA isoforms (PORE-cupine)8. In the present study, we further optimized SHAPE chemical probing and used signal alignments from direct RNA sequencing to identify RNA modifications along a single RNA molecule. We also tested different clustering strategies to group modified RNA reads into their corresponding structures. We named our single-molecule RNA structure probing method ‘sm-PORE-cupine’. Using sm-PORE-cupine, we show that we can accurately separate riboswitch structure ensembles, identify RNA structure ensembles along different SARS-CoV-2 subgenomic RNAs (sgRNAs) and determine structural heterogeneity in the C. albicans transcriptome. Our results demonstrate that RNAs are highly heterogeneous in transcriptomes and that changes in structural ensembles can influence gene regulation inside cells. Results High modification rates and low false-positive rates enable accurate clustering of RNA structure populations To identify paired and unpaired bases along an RNA, we previously sequenced RNAs that are modified and unmodified with the SHAPE compound NAI-N3 (ref. 9), using direct RNA sequencing8. We observed that NAI-N3-modified bases can shift the current mean and standard deviation during direct RNA sequencing, as compared to unmodified bases, to enable us to provide an aggregate structure signal across many molecules for a particular RNA. As the same RNA can fold into different conformations and direct RNA sequencing enables one molecule to thread through a pore at a time, we were curious whether we could separate structure populations based on their modification pattern per molecule (Fig. 1a). As many factors could influence our ability to cluster RNA modification patterns along each molecule, we performed simulation experiments using synthetic reads to determine the impact of structural similarity, read depth, read length and modification rate on our ability to form structural clusters accurately (Extended Data Fig. 1a). We observed that increasing the sequencing depth, length and modification rate of the reads all improved our ability to separate structural populations and that a modification rate of >1.5% coupled with longer read lengths of >750 bases can separate structural populations at a sequencing depth of 1,000 reads per transcript (Extended Data Fig. 1a). Additionally, as different ways of identifying SHAPE modifications using direct RNA sequencing, such as using PORE-cupine or Tombo10, could have different sensitivities and false-positive rates (FPRs) in detecting structure modifications, we also tested the effect of FPR on our ability to separate synthetic RNA populations (Extended Data Fig. 1b). We observed that high FPR decreases our ability to cluster RNA structure populations. To determine the most important parameters out of all tested, we performed linear regression on them and identified that high modification rates and low FPRs have the biggest impact on our ability to separate RNA structure populations accurately (Fig. 1b). Fig. 1 | Experimental and analytical workflow of sm-PORE-cupine. a, Schematic of the experimental workflow to determine RNA structure ensembles from single-molecule RNA structure probing data using nanopore direct RNA sequencing. b, Bar chart shows the effects of each parameter (modification rate, read length, read depth, similarity and FPR) in ena (...truncated)