Identification of conserved RNA secondary structures in human respiratory syncytial virus A (hRSV A) Authors: Damilare K Dabiri, Warren B Rouse, Walter N Moss Published: 2025 ABSTRACT Abstract Respiratory Syncytial Virus Type A (RSV A), a major cause of lower respiratory tract infections in vulnerable populations, is a nonsegmented, negative-sense RNA virus with a 15-kilobase genome encoding eleven proteins. RNA secondary structures may be critical for RSV processes like transcription, translation, and genome packaging. This study used a variety of computational tools, including ScanFold and RNAfold, to identify 123 thermodynamically stable RNA structures in the positive-sense antigenome and 113 in the negative-sense genome. Of these 103 and 98 showing structural conservation, respectively. Analysis revealed differential ti/tv ratios between stems and loops, suggesting RNA structural constraints on RSV evolution. Key structures were found in the L (RdRp), Fusion (F), Nonstructural 1 and 2, Matrix 2 (M2), Matrix (M) genes, and the 3′ UTR (155 nt Trailer sequence). These findings highlight potentially functional RNA structures that may regulate viral replication, offering novel targets for antiviral drug development. Introduction Respiratory Syncytial Virus (RSV) is an enveloped, negative sense nonsegmented, single-stranded RNA (ssRNA) virus. Initially classified under the Pneumovirinae subfamily of Paramyxoviridae, it was reclassified into the Pneumoviridae family (genus Orthopneumovirus), alongside human Metapneumovirus (HMPV) in 2016 [1]. RSV comprises two antigenic types, RSV-A and RSV-B, differentiated by variations in the attachment (G) glycoprotein. Both types co-circulate globally, with RSV-A often linked to more severe disease. It primarily causes lower respiratory tract infections (LRTIs), including bronchiolitis and pneumonia, in infants, young children, older adults and immuno-compromised patients, often the leading cause of infant hospitalization globally. Severe outcomes occur in premature infants, immunocompromised individuals, and those with chronic cardiopulmonary conditions. Recent advances in maternal vaccines (e.g. Abrysvo) and monoclonal antibodies (Beyfortus) aim to reduce this burden [2, 3]. However, RSV’s ubiquity, clinical impact, and evolving subtypes highlight the need for improved prevention and equitable healthcare access. As is typical for RNA viruses, their genetic material is exclusively composed of RNA, which directs essential infection processes, including replication, transcription, translation, and virion assembly within host cells. These mechanisms are often mediated by functionally critical secondary structures found in the viral RNA. Similar to RNA in higher organisms, viral RNA has the potential to form helical structures whenever two complementary sequences are present. These structures offer a strategy through which the virus carries out its biological activities. Different functional RNA secondary and tertiary structures have been identified across RNA viruses [4, 5, 6]. Some notable recent examples are an RNA hairpin structure in the capsid coding region (cHP) of Dengue virus directs translation start site selection in human and mosquito cells [7], and an internal ribosome entry site discovered in Hepatitis C virus that orchestrates cap-independent translation initiation in host cells [5]. Additionally, conserved RNA secondary structure has been identified across the genome of Influenza A virus [8], which can serve as potential drug targets for small molecules, (providing druggable pockets) [9] or for antisense approaches [10]. Preliminary studies of RSV [11] highlighted the potential for forming thermodynamically stable RNA structures. Two regions within the F and M2 genes were modelled to contain highly structured RNA folds. The M2 mRNA transcript is translated into two distinct proteins, M2-1 and M2-2, through a coupled translation mechanism that remains poorly understood. The presence of the conserved secondary structure in this region strongly suggests its role in facilitating the coupled translation process [4]. Additionally, reports have indicated the presence of an RNA secondary structure at the 3′ end of the RSV antigenome, which plays a role in the replication process [12]. Overall, these findings highlight the potential of RNA secondary structures in regulating key aspects of RSV biology and gene expression leaving out other potential structures in the other genes and the genome (–RNA) of the virus. These encouraging results suggest additional (and novel) analyses may reveal additional RNA structural elements that can be playing important roles in this virus. To gain deeper insight into the roles of RNA secondary structure in RSV, our study uses a genome-wide approach that integrates four complementary methods: thermodynamic energy minimization, comparative, and covariation analysis, the examination of transition-to-transversion rates in stem and loop regions, and mapping of identified structures to functional regions in viral genes. Using this novel pipeline, we systematically explored RNA structural features in human RSV (hRSV). This integrated approach allowed us to highlight 11 RNA structures in the RSV positive strand (antigenome) and 3 in the negative strand (genome) that are of particular interest. In this context, negative-sense (–RNA) refers to the viral genome of RSV. Since the RSV genome is negative-sense, it cannot be directly translated into proteins by the host cell’s ribosomes. Instead, the negative-sense genome must first be transcribed into subgenomic, capped, and polyadenylated messenger (m)RNAs that correspond to each viral gene. These mRNAs are then used as templates for protein synthesis. Additionally, during viral replication, a full-length positive-sense RNA, referred to as the antigenome (cRNA), is synthesized from the genome. This antigenome serves as a template for the production of new negative-sense genome (vRNA), which is then packaged into progeny virions. To avoid confusion, in this work, negative-sense (vRNA) will be used interchangeably with “genome,” while positive-sense (cRNA) will be referred to as “antigenome.” This distinction ensures clarity when discussing the different RNA species involved in the RSV replication cycle. Our analysis revealed potential functional RNA structures that overlap with critical RSV genes. These findings provide a detailed roadmap for future RSV research and pinpoint local secondary structure with strong predictive metrics which are supported by phylogenetic evidence and also highlights structures of likely functional importance within the viral genome Materials and methods Data set The complete reference genome of human Respiratory Syncytial Virus Type A (hRSV-A; NCBI accession M74568.1) was retrieved from the National Center for Biotechnology Information (NCBI) nucleotide database. The sequence obtained corresponds to the positive sense (+RNA) form, also known as the antigenome, and serves as the replicative intermediate during the viral lifecycle. To generate the genomic RNA (–RNA) form, the functional negative-sense RNA genome, the reverse complement of the +RNA sequence was generated computationally (https://rna2dmut.bb.iastate.edu/seqrc). This approach ensured the availability of both positive sense (antigenome, +RNA) and negative-sense (genome, –RNA) sequences for downstream comparative and functional analyses. Thermodynamic structural analysis using ScanFold The positive and negative strand of RSV were analyzed for sequence-ordered thermodynamically stable regions using the ScanFold pipeline. The ScanFold tool consists of two main scripts: ScanFold-Scan and ScanFold-Fold, available at https://github.com/moss-lab/ScanFold. ScanFold-Scan scans the sequences using a sliding window approach (120 nucleotide [nt] window size; 1 nt step size). For each window, the Gibb’s minimum free energy (MFE) of folding (ΔG) and corresponding secondary structure were predicted using RNAfold [13]. To evaluate propensity for sequence-ordered (potentially evolved) thermodynamic stability, a Z-score was computed by comparing the native sequence’s ΔG to the mean ΔG estimated for randomized sequences (with same nt content) using the formula: where σ represents the standard deviation of randomized sequence ΔGs. Negative Z-scores quantify the native sequence’s enhanced stability relative to randomized variants (expressed in standard deviations). A P-value was calculated based on the percentage of randomized sequences that showed higher stability compared to the native sequence, acting as a confidence measure (with lower P-values indicating greater reliability). Structural heterogeneity was evaluated using ensemble diversity (ED), which was determined through partition function analysis to measure conformational variability. Lower ED values indicate a predominant single conformation, whereas higher values imply conformational diversity or lack of a stable fold [14]. ScanFold-Fold utilizes Z-score calculations to generate unique Z-score weighted consensus secondary structures that compile the most stable base pairs across all scanning windows: i.e. paired nucleotides that consistently appear in low Z-score analysis windows—highlighting recurring and stable structural patterns [15]. Specifically, all potential competing pairs for a given nt are evaluated using a Znorm metric. This metric considers all predicted interactions for each nucleotide. To calculate Znorm, the Zsum is first determined by summing all instances of a nucleotide participating in base pairs. This Zsum value is then divided by the number of analysis windows in which the nucleotide is paired, yielding the Znorm [14, 15]. This produces a coverage-normalized Z-score that prioritizes base pair arrangements consistently found in low Z-score windows, offering a standardized metric for comparing regions with limited window coverage, such as near the sequence ends where nucleotides are included in only a few windows. The pairing with the lowest Znorm is identified as the most stable and favorable configuration, which is then used to construct the consensus structural model [15]. These structures favor organized stability and probable functionality. All structures with at least one base pair exhibiting a Z-score of −1 or lower are selected for subsequent analyses, as used in previous studies. This threshold is chosen to focus attention on the base pairs with the greatest degree of ordered stability, and which were previously found to correlate with higher prediction accuracy [16]. Metrics derived from ScanFold encompass the MFE ΔG° (indicating thermodynamic stability), Z-score (reflecting ordered stability that may suggest functional roles), ED (assessing the conformational variability of predicted structures), and a P-value (serving as a quality assurance metric). All structure identifiers, transcript coordinates, ensemble ID, sequence data, in silico ScanFold structures, RNAfold re-evaluated ScanFold structures, structure/sequence length, GC content, and positions within the transcript for both positive and negative strands are available in Supplementary File S1. Comparative analysis All structures with a Z-score < -1, identified in both the positive and negative strand of hRSV (105 and 123 structures, respectively), were examined for conserved structure and covariation using the cm-builder Perl script. This script leverages the RNA Framework toolkit and incorporates Infernal (version 1.1.2) [17, 18] to construct and identify covariance models for the structures predicted by ScanFold [19]. This analysis aimed to identify conserved and functionally relevant structural elements by detecting compensatory base pair changes across sequences. To create a database for infernal, sequence data collection was obtained from Bacterial and Viral Bioinformatics Resource Center database (BVBRC) [20]. The initial search focused on the family Pneumoviridae, which consists of the Metapneumovirus and RSV genus. This set of sequences was selected to balance alignment quality with the need for diverse sequences for a robust covariation analysis: i.e. through an expanded sequence pool that contains closely related genera. The search parameters were further refined by applying filters based on (i) genome completeness and (ii) the exclusion of sequences from the “lab host group.” This refinement led to the identification of 8203 Pneumoviridae sequences, which were subsequently downloaded and consolidated into a single batch file containing all the sequences. To identify appropriate sequences for the Infernal database, a systematic pairwise sequence alignment approach was implemented using EMBOSS needle pairwise alignment [21]. Here, the hRSV reference genome served as query sequence against the BV-BRC database sequence serving as the target sequence which was followed by a critical selection criterion to include only sequences with pairwise identity percentages between 60% and 80% [22]. This range was specifically chosen to optimize the detection of sequence variations while maintaining reliable alignments and structure. This careful balance in sequence identity is essential because extremely high sequence identity (>80%) can mask subtle structural variations, while very low identity (<60%) may lead to unreliable structural alignments and potentially false positive predictions. The methodology employed here reflects a strategic approach to database construction, ensuring that the selected sequences contain sufficient evolutionary diversity to support meaningful structural analysis while maintaining reliable sequence relationships. In total, of the 8203 sequences downloaded from BVBRC, 5304 sequences had 60%–80% sequence identity to the reference sequence and were used in creating the infernal database. The structural alignment files, formatted in Stockholm format, were examined for covarying base pairs and further analyzed using the CaCoFold algorithm and R-scape (version 1.5.16). Statistical significance was assessed using the APC-corrected G-test [23, 24], with a default E-value threshold of 0.05 to determine the reliability of the detected covariation patterns [23]. For clarity, covarying base pairs in RNA secondary structure refer to pairs of nucleotide positions in an RNA molecule that undergo complementary mutations across different sequences while maintaining base pairing. This phenomenon is strong evidence for functional RNA secondary structures, as it indicates evolutionary pressure to maintain structural stability [25]. Since the database focused on the family Pneumoviridae, which consists of the Metapneumovirus and RSV genus, a biopython script was developed to highlight the different numbers of various species that makes up each alignment. This Python script processes Stockholm files generated from the Infernal algorithm for each conserved structure. It performs two main tasks sequentially. First, it updates the Stockholm files (.sto) by appending NCBI definitions to valid accession numbers, fetching definitions from the NCBI nucleotide database using BioPython’s Entrez module and validating accession numbers with a regex pattern. Second, it extracts summary statistics from the updated files, categorizing species (e.g. RSV A, RSV B, Bovine RSV, Human Metapneumovirus, and others) based on keywords in definitions. It counts species occurrences, extracts motif numbers for sorting, and compiles counts into a tab-separated output file with a header. The “Others” group consist of mainly hRSV that are neither designated as Type A or B (UNVERIFIED) in the accession number definition. To further evaluate and identify structural sequences in the negative-sense RNA that overlap with the positive-sense RNA, a Python script was developed to perform a comparison between negative-sense and positive-sense sequences. The script processes +RNA and −RNA sequences from text files, generates reverse complements of the −RNA, and detects overlaps—either exact matches or substrings with the +RNA sequences. This analysis helps determine whether conserved −RNA structures are reverse complements of +RNA structures, thereby addressing potential redundancy in RSV structural studies. The python scripts can be found on https://github.com/moss-lab. The pairwise sequence alignment result, Stockholm alignments, R-scape/CaCoFold results, species count in stockholm file and output from the overlapping analysis can be found at Supplementary File S2. Estimation of transition versus transversion in stem regions Structures identified through homologous sequence alignments, alongside known structures displaying covarying base pairs in the positive and negative strands (103 and 98, respectively) through sequence alignment and conservation analysis were further examined by estimating the transition/transversion (ti/tv) rate in the stem region. This approach aims to identify evidence of selection in stems that lack sufficient variability to deduce statistically significant covariation. Essentially, we perform this analysis to help rank ScanFold structures and extract additional targets of interest beyond those that have strictly selected covarying base pairs. Non-covarying structures may manifest selective pressure through biased nucleotide substitution patterns. High ti/tv ratios in stems, where transitions preserve Watson–Crick pairing more frequently than transversions, provide evidence of structural constraint and suggest functionality. To achieve this, a modified Biopython script, adapted from Madhumista et al. [26], was used. In brief, this script analyzes RNA sequence conservation and mutation patterns relative to secondary structure. Input data included a FASTA file of aligned RNA sequences adapted from the Stockholm file (.sto) of each conserved structure from the previous step and its corresponding dot-bracket file defining the secondary structure (stems: () pairs; loops:). Sequences were converted to RNA (T→U). A consensus sequence was generated using per-position majority pattern which is definitely the same file nucleotide pattern as seen in the dot bracket annotation file. Structural annotations from the dot-bracket notation were integrated to categorize residues as stem- or loop-associated. A conservation table was constructed to report nucleotide frequencies (A, U, G, C) at each position, alongside the consensus base and structural annotation. Mutation frequencies were quantified by comparing aligned sequences to the consensus, distinguishing transitions (A↔G, C↔U) from transversions. For each mutation type (e.g. A→G), counts in stems (MCounts_S) and loops (MCounts_L) were normalized by the total occurrences of the consensus base in each structural context (Count_S, Count_L) to calculate normalized frequencies (NFrequency = MCounts / Count). Transition-to-transversion (ti/tv) ratios were computed as the sum of normalized transition frequencies divided by the sum of normalized transversion frequencies for stems and loops separately. All analyses were implemented in Python using Biopython and pandas, with outputs saved as CSV tables. The script proceeds to generate a consensus table summarizing nucleotide frequencies (A, U, G, C) across positions, calculate mutation statistics, and classify mutations as transitions or transversions. The script also computed transition-to-transversion (ti/tv) ratios and performed Spearman correlation analysis to compare mutation frequencies in stem and loop regions. To identify biologically relevant structures, two criteria were applied: (i) structures with a Spearman correlation >0.5 and statistical significance (P < 0.05) AND (ii) Structures where the stem ti/tv ratio exceeded the loop ti/tv ratio. This analysis is grounded in the observation that while the expected ti/tv ratio is ∼0.5 (as transversions are more likely), observed ratios often exceed 2.0, indicating greater tolerance for transitions. Transversions are more likely to destabilize RNA secondary structures, whereas mutations in loops have minimal impact, except in tetraloops [26]. These findings suggest that structural and functional constraints favor transitions in stems, highlighting evolutionary and functional significance. This approach ensures the selection of structures with meaningful mutational patterns and evolutionary constraints in stem regions. The biopython script can be found on https://github.com/moss-lab. The result and generated CSV file can be seen in the Supplementary File S3. Genomic localization of relevant secondary structures RNA secondary structure identified in the analysis of transition/transversion (ti/tv) ratios in stem regions—19 and 20 in total—were mapped to their corresponding sites on the RSV genome. These coordinates were cross-referenced with gene annotations using the Uniprot Data base [27] and prior RSV gene characterizations to assess overlap with regions linked to essential viral processes, such as replication, translation, or immune evasion. This step aims to link structural conservation with functional relevance. Structures residing in regions experimentally validated or computationally predicted to mediate viral infectivity, genome packaging, or host interactions were regarded as potential functional elements. The positive-sense (+) RNA of RSV, unlike the negative-sense (–) RNA, aligns closely with mRNA-based annotations in UniProt, enabling direct genomic mapping and functional interpretation. Thus, our final analysis prioritizes +RNA to assess conserved RNA secondary structures Overall, to assess the functional significance of RNA secondary structures, we employed a comprehensive four-pronged bioinformatics approach using: (i) sequence-ordered thermodynamic stability, (ii) conservation and covariation analysis, (iii) transition/transversion (ti/tv) substitution bias in helical regions, and (iv) genomic localization of structures which is illustrated in Fig. 1. This multi-dimensional approach enabled the identification of candidate secondary structures potentially involved in essential viral processes such as genome packaging and host–pathogen interactions. High-confidence structures were prioritized based on their putative contributions to viral fitness, providing a foundation for future experimental validation of their mechanistic roles in RSV biology. A visual representation of the materials and methods utilized in this study. (A) The hRSV positive sense (+RNA) was retrieved from NCBI and converted to its corresponding negative sense RNA (–RNA) by reverse complementation using the Reverse Complement tool. (B) Both the positive (+RNA, antigenome) and negative sense (–RNA, genome) sequences were analyzed using ScanFold to predict thermodynamically stable RNA secondary structures (Z-score ≤ –1). This sliding-window approach evaluated MFE (ΔG), and Z-scores to identify structured regions. (C) Thermodynamically stable structures identified in ScanFold were assessed for evidence of evolutionary covariation using cm-builder. (D) Conserved stem-loop structures were further analyzed by calculating transition/transversion (ti/tv) ratios. In the panel an example alignment is annotated with transitions (ti) and transversions (tv) in green and red, respectively (with respect to SEQ1). (E) Significant structures identified in output from the ti/tv analysis were mapped to genomic locations and cross-referenced with UniProt annotations to infer biological significance. Results and discussion Genome-wide RNA structural profiling of RSV using ScanFold A comprehensive exploration of RNA secondary structure within the RSV genome was conducted using the ScanFold pipeline to analyze both the positive (antigenome/mRNA) and negative (genome) strands. This involved employing a sliding window approach, utilizing a 120 nucleotide (nt) window with 1-nt step increments. This process generated 15106 overlapping windows, which provide insights into the structural characteristics of these viral RNAs: the resulting data are visualized/summarized using the Integrative Genomics Viewer (IGV) [28] (Supplementary File S1). Computational predictions of MFE and thermodynamic Z-scores were calculated for each window to assess structural stability and potential selective pressures for sequence-ordered thermodynamic stability. The antigenome exhibited MFE values ranging from −35.3 to 0.0 kcal/mol, coupled with average Z-scores spanning −3.36 to 1.00, indicating a nuanced variability in structural stability, while the genome displayed a broader range of MFE values (−35.4 to −4.5 kcal/mol) and Z-scores (−4.4 to 2.5) (Table 1). Although the genome exhibited greater average thermodynamic stability overall, as indicated by its lower MFE, the antigenome is predicted to have a higher propensity for sequence-ordered thermodynamic stability, as evidenced by its more negative average Z-score (−RNA: −0.44 versus +RNA: −0.56) and an increased proportion of structures exhibiting a Z-score < −2 (−RNA: 7.24% versus +RNA:9.10%). Table 1.ScanFold Metrics of RSV positive strand (+RNA) and negative strand (−RNA)StrandGCMFE range (kcal/mol)Avg_MFE Z-score RangeAvg Z-score% nt Z-score <−2% nt Z-score < −2Structures < −2 Z-scoreStructures < −1 Z-scoreED −RNA 33.2−35.4 to −4.5−18.8−4.4–2.5−0.457.20%58.80%1611327 +RNA 33.2−35.3 to 0.0−15.4−3.3–1.0−0.569.10%60.40%2112324 ScanFold Metrics of RSV positive strand (+RNA) and negative strand (−RNA) The −RNA alongside the + RNA which remain persistently encapsidated by nucleocapsid (N) proteins throughout the viral life cycle, experience structural destabilization due to N protein interactions, which disrupt potential RNA secondary structures within the genome. In contrast, mRNA, distinguished from +RNA by the presence of a 5′ cap and polyadenylated tail, appears to be under stronger selective pressure to preserve functionally relevant RNA structural elements. This is in line with studies on Influenza A virus (IAV), where regulatory elements critical for replication and translation are preferentially located in the (+) RNA [29]. This RNA structural bias seen in the RSV positive strand is also likely reflecting the additional biological pressures on the RSV + RNA to undergo post-transcriptional processing, cellular transport, and translation. In the −RNA, no distinct RNA structures were detected within the 3′ UTR 44 nt leader sequence and the 5′ 155 nt trailer sequence. However, the +RNA contained a stable structure in the 3′ 144 nt trailer sequence, characterized by a Z-score of −3.07 and an MFE value of −15.7 kcal. The absence of stable structure at the 3′ terminus of the +RNA may be related to the need for RdRp attachment to initiate transcription and replication as studies have shown that the 1−11 nt in the leader sequence are required for initiation of transcription [30]. Conversely, according to the study by Sarah et al., the structure observed in the +RNA trailer sequence may be involved in the polymerase backtracking mechanism required for the synthesis of the −RNA strand [12], although the structure identified in this current study is located upstream of the previously described 25-nt segment. Aside, this structure in the 3′untranslated regions of RSV, multiple additional RNA structures identified in this study are located within the RSV coding regions. Another important metric from the ScanFold output is the ED, a prediction of the heterogeneity of secondary structures an RNA molecule can adopt. Generally, a higher ED value indicates greater conformational plasticity, predicting that the RNA can fold into a wider range of structures with similar free energies. Conversely, a lower ED suggests a more compact and stable structural ensemble, which is often an evolved characteristic of functional RNAs [31, 32]. The average ED value varies across different RSV genes and between their positive (+) and negative (−) strands ​. Specifically, the negative-sense strand (antigenome) exhibits a higher average ED of 27, compared to the positive-sense strand (genome) with an average ED of 24. This heterogeneous state of structures may be indicative of the multiple roles they play in the viral lifecycle. Parallel studies have reported RNA structural polymorphism in HIV-1, a ssRNA virus, which heavily relies on alternative RNA folding to coordinate its life cycle [33]. Similarly, SARS-CoV-2 RNA exhibits secondary structure heterogeneity, with alternative conformations being critical for promoting frameshifting rate [34]. A total of 21 RNA secondary structures were predicted with average Z-scores < −2 (Zavg < −2) across the +RNA sequence. These structures were dispersed across eight gene segments, which are N (Nucleoprotein), P(Phosphoprotein), M (Matrix Protein), SH (Short Hydrophobic Protein), G (Glycoprotein) F (Fusion Protein), M2 (Matrix 2 Protein), and L protein (Large Protein, RDRP), with one structure identified at the 3′ UTR (155 nt trailer sequence). Analysis of the −RNA revealed 16 distinct structures with < -2 Z-scores. These structures were identified in 6 of the 11 genes, including the N, M, SH, F, M2, and L (Table 2). To better understand the distribution of these structures, a breakdown of each gene’s MFE and Z-score was done for both the +RNA and −RNA strand, summarized in Table 2. The distribution of structures across the genome, alongside the folding metrics and dot bracket annotation can be found in Supplementary File S1. Table 2.Distribution of the Z-score across the genes in the −RNA and +RNA strand Zavg% nt < −2 Zavg% nt < −1 Zavg% nt < 0 ZavgNo. of structures Zavg < −2No. of structures Zavg < −1EDGenes+−+−+−+−+−+−+− NS1 −0.59−0.313.1818.1114.7939.7071.5472.09004527.9126.16 NS2 −0.40−0.358.710.0028.710.9944.3524.55005224.8331.36 N −0.68−0.5310.713.4232.8326.7576.2958.511291024.9326.42 P −0.65−0.6712.458.4138.6538.5469.3272.82109822.1524.32 M −1.40−0.6523.331.8763.5415.6393.8566.5663161119.3628.12 SH −1.10−0.3418.908.0057.0433.4984.9552.91118114.3228.84 F −0.75−0.6517.328.5638.63 26.03 69.92 71.54 6 4 9 524.3628.28 G −0.76−0.3112.541.2937.52 14.59 83.46 53.87 1 0 15 1715.7228.79 M2-1/2 −1.24−0.9713.1012.1163.05 20.38 79.33 56.31 2 1 7 1223.3327.32 L −0.31−0.313.108.2923.68 31.58 62.91 69.34 3 4 41 4026.7727.45 Total Avg −0.80−0.6012.336.7061.84 58.37 25.82 34.94 2 1 12.3 1.122.3727.71 Distribution of the Z-score across the genes in the −RNA and +RNA strand Our analysis of Z-scores across the RSV genome reveals significant variation in RNA secondary structure potential between genes and between the positive and negative strands (Table 2). These differences suggest specialized structural requirements for various viral processes, with particularly stable structures appearing in coding strand. Analysis of Z-score distributions across the positive strand of the RSV genome found average Z-scores ranging from −0.31 (L gene) to −1.40 (M gene). This spectrum indicates varying propensities for forming sequence-ordered secondary structures across different genomic regions. Genome-wide, 61.84% of positive strand nucleotides show Z-scores < 0, indicating a general tendency toward structure formation. The M (93.85%), SH (84.95%), and M2-1/2 (79.33%) genes contain particularly high proportions of nucleotides ordered for forming stable structures. Notably, 12.33% of positive strand nucleotides have Z-scores < −2, representing especially stable and ordered structural elements. The M gene contains the highest concentration of these extremely stable sequences (23.33% with Z-score < −2), followed by SH (18.90%), F (17.32%), and M2-1/2 (13.10%). Structural analysis reveals distinct patterns, with M and F genes each containing six highly stable (Z-score < −2) structures. The L gene shows an abundance of moderately stable structures (41 with Z-score < −1), consistent with its multiple functional domains. Overall, the identification of such stable structures in the RSV positive-strand RNA suggests regulatory functions, potentially including regulation of translation via polymerase pausing or frame-shifting elements, as observed in other paramyxoviruses [29]. The negative strand on the other hand, displays generally less pronounced structural potential, with average Z-scores ranging from −0.31 (L and NS1 genes) to −0.97 (M2-1/2 gene). This trend may reflect the strand’s primary role as a replication template and transcription substrate, which does not require the same level of regulatory control as coding RNA. While 58.37% of negative strand nucleotides show Z-scores < 0, the distribution varies substantially between genes. The P (72.82%) and M (66.56%) genes show high structural propensity, whereas NS2 (24.55%) exhibits the lowest. Only 6.70% of negative strand nucleotides reach Z-scores < −2, with M (18.30%) and P (16.21%) genes containing the most stable structures, a contrast to their positive strand counterparts. Ordered structures in the positive-strand RNA leave structural “echoes” in the reverse-complemented regions of the negative-strand RNA, contributing to these observed values. Identification of conserved RNA structures in RSV To help prioritize structures of particular interest we assessed the evolutionary conservation of ScanFold secondary structure models. Here, we generated covariation models for all ScanFold predicted structures using Infernal. Models were built from a multiple sequence alignment of >5000 RSV genomes (as explained in the Materials and methods; all data are presented in Supplementary File S2). Statistically significant base-pair covariations were identified using R-scape’s G-test: a summary of the result can be seen in Table 3. For the positive strand (+RNA), Infernal identified 103 structures as being conserved across RSV genera. Of these conserved structures 9 contained statistically significant covarying base pairs, indicating evolutionary conservation of structure sufficient to observably affect RSV genome evolution. The L gene had 35 conserved structures, with 5 further supported by covariation. The trailer region (TrC) contained one conserved structure with a covarying base pair, indicating evolutionary pressure at the 3′ UTR end. While conserved structures were predicted in M2-1/M2-2 (Table 2), none were supported with statistically significant covariation. Table 3.Distribution of conserved and structures with covarying base pairs across the genes in the +RNA and −RNAAntigenome (+RNA)Genome (−RNA)RegionConserved structuresCovarying structuresConserved structuresCovarying structures NS1 4021 NS2 4030 N 9160 P 7060 M 170101 SH 5010 F 141152 G 4041 M2-1/2 30142 L 355376 TRc 1100 Distribution of conserved and structures with covarying base pairs across the genes in the +RNA and −RNA The RSV genome (−RNA) yielded 98 structures with evidence of conservation, where 13 contained R-Scape identified covarying base pairs. Of these 98 conserved −RNA structures, 51 overlap with the sites of reverse complemented +RNA structures, where 27 (35.5%) substantially overlapped (>90% sequence) +RNA structures (Supplementary File S2). While many of the conserved structures predicted in the -RNA are likely “echoes” of + RNA structures, we did predict 25 unique −RNA structures with no + RNA overlap. Of these, 10 had support from statistically significant sequence covariation indicating that selective pressure may be working to maintain these structures for some −RNA functions. While the −RNA shows less overall propensity for ordered structure, and its tight association with protein is generally unfavorable for structure formation, such small local structures may still be able form: e.g. during transient stages when fragments of “naked” −RNA are present. Furthermore, the analysis of evolutionary depth and species distribution reveals that the majority of sequences contributing to conservation and covariation metrics originate from hRSV subtypes A and B, bovine respiratory syncytial virus (BRSV) and other hRSV variants without assigned types as shown in Supplementary File S2. Notably, human metapneumovirus (hMPV) representatives were not identified across the conserved structural elements. This distribution is consistent with the close phylogenetic relationship between hRSV and BRSV, which likely stems from shared ancestral origins and similar genetic architectures [35]. This strong phylogenetic link between hRSV and BRSV is key to identifying conserved RNA motifs, as evolutionary pressures likely preserve functional RNA structures critical for virus lifecycle in both species. The consistent presence of these structures in hRSV and BRSV, coupled with their absence in hMPV, highlights the specificity of these conserved motifs to the RSV genus. This suggests that while Pneumoviridae share broad characteristics in their biology, the identified RNA structures are unique to RSV’s evolutionary and functional needs. The observed pattern reinforces the robustness of covariation-based approaches in pinpointing functionally significant RNA regions by detecting coordinated changes in paired nucleotides within evolutionarily conserved structures, thereby highlighting their critical roles in the viral lifecycle. Transversion and transition in stem versus loop regions As an orthogonal approach to evaluate evolutionary constraints of RNA structure and to help prioritize targets, we employed a strategy based on the rate of transition and transversion mutations within the stems and loops of these structures. This method can identify structures that exhibit evolutionary constraints, as indicated by specific mutation patterns, which nonetheless may not have sufficient variation for identification of covariation. By leveraging transition and transversion rates, this comparative approach allows for the identification of additional RNA structures with potential functional importance. This approach, as detailed in the Materials and methods section, identified 19 structures in the +RNA exhibiting a positive correlation between the stem transition/transversion (ti/tv) ratio and a stem ti/tv ratio greater than that of the loop in the antigenome, as summarized in Table 4 (Supplementary Folder S3). These findings overlap with the presence of covarying base pairs in key structures, such as structure 104 (nt 12897–13 006), structure 112 (13904–13917) structure 119 (14706–14771), all in L gene, and structure 123 (15048–15104) in the Trailer sequence (Fig. 3). These results highlight the likely functional significance of these structures, where multiple lines of evidence point to their importance. This is not to say that the other identified structures are not playing roles; however, these particularly interesting structures can help prioritize addition work: e.g. experimental validation of function or structural biology. Table 4.+RNA structures identified in the ti/tv analysisStructureGeneCoordinatesSpearman correlation P_ValueStem_ti/tvLoop_ti/tvMFE (kcal/mol) Z-scoreCovarying bpSTR_3NS1447–4990.603.75E-021.621.59−10.5−2.94 STR_4NS1502–5350.881.88E-041.281.03−1.9−2.27 STR_9NS21043–10700.872.32E-041.150.99−1.5−0.87 STR_39M3843–38580.763.91E-034.001.17−1.8−3.17 STR _43M4022–40400.691.36E-021.441.30−3.7−2.37 STR_44M4050–40650.671.62E-020.670.25−2.9−3.93 STR_63F6085–61400.614.00E-021.211.20−25.9−4.35 STR_64F6372–64070.594.00E-021.010.50−17.4−4.96 STR_71F7160–72160.872.74E-042.001.53−10.6−1.48 STR_72F7221–72950.938.47E-062.351.86−19.5−3.11 STR_76F7493–75400.845.90E-041.160.95−8.5−3.32 STR_80M28354–84350.821.24E-031.511.40−13.7−2.46 STR_101L12255–12 2970.905.83E-050.780.75−13.2−1.82 STR_103L12787–12 8640.661.86E-024.171.48−11.4−1.97 STR_104L12897–13 0060.829.61E-041.881.67−18.0−2.25+STR_110L13754–13 7970.846.21E-042.711.82−5.3−0.97 STR_112L13904–13 9170.623.22E-027.002.25−3.8−6.10+STR_119L14706–14 7710.613.60E-022.431.54−14.3−1.49+STR_123TRc15110–15 1390.632.71E-023.231.03−4.1−1.35+ +RNA structures identified in the ti/tv analysis The negative-sense RSV genome (−RNA) revealed 20 structures meeting stringent conservation criteria, defined by a Spearman correlation > 0.5 (P < 0.05) and a higher stem transition/transversion (ti/tv) ratio compared to loops. Similar to our previous analysis of conservation/covariation, 9 out of the 20 structures in the −RNAs sequence that showed skewed ti/tv ratios had some partial or complete overlap with +RNA hits, where 2 of these 20 structures were exact reverse complements of +RNA structures (Supplementary File S2). While 11 unique structures with no complementarity were identified for the −RNA structures. These findings suggest that specific −RNA sequences in RSV may form functional structures with strand-specific roles, despite the limited regulatory pressures on −RNA and its nucleoprotein encapsulation in nonsegmented negative-strand viruses (nsNSVs). Detailed conservation and overlap data are presented in Table 5 and Supplementary File S1. Table 5.−RNA structures identified in the ti/tv analysisStructureGeneCoordinatesSpearman correlation P_ValueStem_ti/tvLoop_ti/tvMFE Z-scoreCovarying bpOverlap + RNA STRSTR_11L1686–17860.671.68E-021.351.24−17.0−1.8 STR_14L2032–20760.821.09E-031.841.76−7.9−1.1++STR_23L3083–31810.613.44E-026.001.89−8.8−0.6 STR_36L5391–54830.801.72E-031.991.19−17.5−1.7 STR_40L5898–59860.746.22E-032.251.68−15.8−1.8 STR_43M26732–67720.661.95E-021.400.96−7.6−3.0 STR_44M26775–68010.755.08E-031.901.25−6.0−3.2+ STR_45M26803–68230.681.55E-0210.764.00−3.2−3.4 +STR_47M26885–69270.746.23E-031.050.93−12.5−2.4 +STR_48M26929–69531.001.09E-021.480.93−2.2−1.5 STR_56F7683–7730_0.603.91E-021.230.72−11.2−3.3 +STR_57F7734–77580.613.35E-021.191.00−11.8−0.5 STR_58F7764–77830.681.51E-021.541.50−1.9−1.2 STR_60F7931–79990.792.37E-032.771.68−11.6−3.2 +STR_69F9083–91380.613.52E-021.231.02−2.8−3.1 +STR_73G9725–97690.863.10E-041.140.86−14.3−3.0 +STR_76G10308–10 3740.632.71E-021.451.14−21.6−3.5 STR_80M11023–11 0860.837.50E-041.060.89−19.9−3.6 +STR_82M11359–11 4250.603.93E-021.341.31−15.0−2.3++STR_101P12752–12 8390.604.07E-021.191.04−20.5−2.7 −RNA structures identified in the ti/tv analysis Genomic localization and prioritization of identified structures To prioritize structures of potential functional importance within the RSV +RNA, a series of systematic computational approaches were employed, as outlined in Fig. 2. This analysis pipeline helps to prioritize and highlight the key conserved and stable RNA structures that may play particularly important roles in viral processes. Importantly, in the final selection stage, structures overlapping regions linked to critical RSV biological processes such as replication, translation, or immune evasion were prioritized. These genes are of particular interest due to their direct involvement in viral function and pathogenesis, making them key targets for further investigation. By applying these criteria, we aimed to rank and prioritize target structures for further research. This approach not only refines the selection of structures for experimental validation but also provides a focused framework for investigating their potential roles in RSV biology. Selection criteria for high-value structure identification in positive strand RSV. An overall summary of the locations of highlighted structures of possible importance can be found in Supplementary File S5. Using these criteria (Fig. 2), 11 structures across 6 genes—Nonstructural Protein 1 (NS1), Nonstructural Protein 2 (NS2), Fusion Gene (F), Matrix Gene (M), Matrix Gene 2 (M2-2), and Large Gene (L)—were highlighted. Additionally, one structure partially overlaps the trailer sequence region. In the NS1 gene, the identified structure is Structure 3 (Fig. 3). This structure exhibits a Z-score of −2.94 and a MFE of −10.5 kcal/mol. It forms a hairpin with a stem-loop architecture, a common motif in RNA secondary structures that enhances structural stability and supports potential regulatory functions. Notably, Structure 3 contains, an internal loop spanning nucleotides 481–488 harbors an AU-rich sequence, which constitutes a favorable binding region for RNA-binding protein (RBP) like ZFP36, a member of the zinc finger protein family of RBPs known to associate with AU-rich regions in viral and host mRNAs, thereby influencing mRNA stability [36]. Consistent with this, ZFP36 is predicted to bind this RNA structure according to analyses performed using RBPmap [37]. (Supplementary File S4). Interestingly, this structure is located in the alpha subunit of the NS1 protein, a region has been shown to negatively affect the transcriptional regulation of genes within key signaling pathways, which are crucial for mediating IFN antagonism. Additionally, expression of this gene reduces the expression of host immune response genes [38]. In the NS2 gene, which synergizes with NS1 for immune evasion, Structure 8 (Fig. 3) spans nucleotides 964–988 (24 nt) and has an MFE of −2.2 kcal/mol and a Z-score of −1.98. Notably, Structure 8 coincides with a segment of the NS2 gene implicated in reducing STAT2 levels, which inhibits JAK1/TYK2-mediated activation of STAT1/2, a critical transcription factor for IFN-inducible antiviral genes [35]. Aside from the significantly low Z-score and MFE values of these structures, they displayed low ED values (compared to the genome average): 6.72 and 1.17, respectively. Additionally, the prevalence of sequences from RSV-A, RSV-B, and, the more distantly related, BRSV in the alignments suggests a significant evolutionary pressure to maintain Structures 3 and 8. This indicates that these structures are likely important for the viruses’ function or survival, leading to their conservation across different RSV species. Summary of ScanFold results on RSV (+) RNA. (Top panel) Genome visualization of resulting data using IGV. From top to bottom the tracks represent the location of the genome (in kb); a base pair (BP) track indicating ScanFold predicted base pairs as arcs with blue, green and yellow arcs indicating Z-score cutoffs of ←2, between −2 and −1, and between −1 and 0, respectively; a histogram of the MFE values of each ScanFold window (with bar placed at the first nt of the window) in kcal/mol; a histogram of window Z-score values with read and blue colors indicating negative and positive Z-scores, respectively; a cartoon of the RSV genome where exons are represented with thicker blue rectangles and sequence polarity with white arrows, each gene is labeled above the corresponding exon; finally, the last track indicates the location and extent of ScanFold predicted structures (STR 3 to STR 123) that are have a high degree of ordered stability, evidence of evolutionary conservation, and overlap genomic regions of interest. (Bottom panel) Secondary structure models of the ScanFold structures (Structure 3 to Structure 123) highlighted in panel (A). Here, base pairs are indicated with thick lines connecting nt and are colored with their Z-score values (←2 in blue and between −2 and −1 in green; see legend of figure). Covarying base pairs (identified by R-scape) are highlighted with gray boxes. The first and last nt of each structure are labeled with their genomic location. In the Fusion (F) gene of the hRSV genome, three RNA structures were identified (Fig. 3): Structure 63 (nucleotides 6085–6140) with an MFE of −25.9 kcal/mol and a Z-score of −4.36, Structure 64 (nucleotides 6372–6407) with an MFE of −17.4 kcal/mol and a Z-score of −4.96, and Structure 72 (nucleotides 7221–7295) with an MFE of −19.5 kcal/mol and a Z-score of −3.11. Structure 63 and 64 are interesting because of their proximity to the fusogenic region of the F0. Aside from their conservation and sequence ordered thermodynamic stability, these structures contain interesting structural motifs. Structure 64 is capped with an AACU tetraloop hairpin which can be stabilized through noncanonical interactions or stacking of bases [39, 40] and Structure 63 has an 8-nt buldge on the right stem which may present these nucleotides for RBP interactions [41, 42]. The fusogenic domain is critical for the function of the F protein, wherein the F0 is cleaved at two independent sites for activation of the gene [43, 44]. Structure 72 (7221–7295) in the Fusion (F) gene of hRSV, with an MFE of −19.5 kcal/mol and a Z-score of −3.11(Fig. 3). It has a variety of internal loops and bulges and is positioned in the Heptad Repeat B region of the F gene. This region is important, as it enhances the stability and functional conformation of the fusion protein, essential for viral entry and membrane fusion [45]. These highlighted structures correlate with our previous preliminary work, which also highlighted similar regions in the F gene for RNA structural propensity [11]. In the M2-2 protein, Structure 80 (positions 8354–8435) exhibits an MFE of −13.7 kcal/mol and a Z-score of −2.46, located at the 3′ end of the M2-2 second open reading frame (ORF) (Fig. 3). This structure features a multi-way junction where multiple helices converge, forming a central hub with unpaired or mismatched bases. Structure 80 is located within the 68-nucleotide overlap region between the M2 and L genes in the RSV genome. This overlap, encompassing the M2 gene-end (GE) and L gene-start (GS) signals, creates a complex regulatory landscape that poses unique challenges for protein translation. This positioning suggests structure 80 may facilitate translation of the downstream L gene, potentially through mechanisms like coupled translation of overlapping ORFs [4], frameshifting [46], or termination-reinitiation [47], enabling efficient translation of the overlapping L gene into distinct proteins. A similar mechanism has been reported for the M2 gene, which encodes the M2-1 and M2-2 proteins via coupled translation, facilitated by a secondary structure 150 bases downstream of the M2-1 stop codon [4]. Structure 39 (Fig. 3), spanning coordinates 3843–3858, has a MFE of –1.8 kcal/mol and a Z-score of −3.17, and is predicted to form a stable stem-loop within the M gene. Previous studies have shown that this protein is exported from the nucleus in RSV-infected cells via a Crm1-dependent mechanism, mediated by the c-Nuclear Export Signal (amino acids 194 to 206) [48]. Notably, Structure 39 overlaps with the region of the protein involved in this export process. Three structurally significant RNA motifs were identified within the L protein, each exhibiting notable stability and evolutionary conservation. Structure 104 (positions 12897–13 006) in the connecting domain (CD) has a MFE of −18.2 kcal/mol and a Z-score of −2.25. Structure 112 (13904–13 917) in the methyltransferase domain shows an MFE of −3.8 kcal/mol and Z-score of −6.1. Structure 119 (14706–14 771) in the C-terminal domain exhibits an MFE of −14.3 kcal/mol with a Z-score of −1.49 (Fig. 3). Notably, these three structures alongside structure 123 are the only ones displaying statistically significant covarying base pairs, coupled with a transition/transversion (ti/tv) ratio bias higher than that observed in loop regions. This combination of features highlights their potential functional importance, suggesting these motifs play critical roles in the L protein’s biochemical activities. Structure 123 (positions 15110–15 139, Fig. 3) exhibits a MFE of −4.1 kcal/mol and a Z-score of −1.35. This structure adopts a classic hairpin conformation with a distinctive, noncanonical U-rich hexaloop at its apex, which may play a crucial role in maintaining RNA architecture and function. The stability of such unusual hexaloops has been previously shown to protect RNA from nuclease degradation [49]. Also, the analysis of conservation identified only sequences from hRSV types A and B and no sequences from BRSV. This is expected, as studies indicate that the BRSV genome contains 161 nt at 3′ untranslated region trailer sequence (3′UTR TRc), which is 6 nucleotides longer than that of hRSV [50]. This minor length difference may suggest that the structure identified herein has a functional role specific to hRSV’s viral life cycle. Furthermore, the structure exhibits a low ED value of 6, a stark contrast to the average antigenome ED of 27. This low ED implies a robust RNA secondary structure, with minimal propensity to adopt alternative conformations. ​Such stability and lack of conformational flexibility are often indicative of important functional roles for an RNA element, where structural precision is required for its activity. Investigating the potential RBPs that may interact with this structure using RBPmap, 35 potential RBPs were identified (Supplementary File S4). Among these 35 predicted RBPs, FUBP3 (far upstream element-binding protein 3) stands out as particularly noteworthy due to its established roles in RNA virus life biology. Previous research has demonstrated that FUBP3 binds to the 3′ untranslated region (UTR) of the Japanese Encephalitis Virus (JEV). This interaction is crucial for promoting viral replication complexes, thereby accelerating viral RNA replication [51]. Although Structure 123 is located within the 3′ untranslated region (3′UTR, 155 nt TRc) and positioned upstream of a previously characterized 25-nucleotide segment known to be involved in polymerase back-priming during RSV genome (−RNA) synthesis [12]. Its stability, significant evolutionary conservation, exceptionally low ED and pronounced transition/transversion (ti/tv) bias collectively emphasizes its potential functional importance. These intrinsic features, coupled with its predicted interactions with RBPs known to engage with similar RNA viruses, strongly suggest a key role for Structure 123. Among the eleven –RNA structures identified by ti/tv analysis which lack complementarity with the +RNA, Structure 44 (coordinates 6775–6801) (Fig. 3) stands out. It has MFE of –6 kcal/mol and a Z-score of –2.97 and also shows to have statistically significant covarying basepair and elevated transition/transversion (ti/tv) ratios in its stem region, indicating conserved base-pairing under evolutionary pressure. Located on the –RNA strand within the M2 gene. The overall analysis for the −RNA can been seen in Supplementary Files S1-S3. This study primarily focuses on two positive-sense RNA species: the antigenome +RNA, which is complementary to the genomic RNA, and viral mRNA. Although these RNAs share synteny, they possess distinct biochemical characteristics. The +RNA lacks essential post-transcriptional modifications such as the 5′ cap and poly-A tail and is encapsidated by the nucleocapsid (N) protein, whereas viral mRNA is not encapsidated by the N protein [26]. Since both are positive-sense RNAs, this study prioritizes the identification of RNA secondary structures within these species, as such structures mediate the majority of viral biological activities like replication and translation. However, this focus does not dismiss the potential significance of secondary structures in the genomic RNA (−RNA). The −RNA is also encapsidated by the N protein, which shields it from host ribonucleases and prevents intermolecular aggregation—an interaction thought to melt RNA structure [27]. Interestingly, a recent study by Wenjie et al. proposed that viral genomic RNA may exist transiently in an N protein-free state within ribonucleoprotein complexes (RNPs). These “naked” states could allow the formation of higher-order secondary structures crucial for selective RNP recognition during virion assembly, as observed in certain orthomyxoviruses, which are segmented, negative-sense viruses [28]. This suggests that secondary structures in −RNA may play important functional roles even within an encapsidated environment. Building on this premise, we highlight selected RNA structures in −RNA that may contribute significantly to viral processes, emphasizing their potential role in the viral life cycle despite the protective encapsidation by N protein. Conclusion In this study, we employed a computational strategy combining four independent yet complementary RNA structure prediction approaches to systematically identify functionally significant RNA secondary structures in the RSV genome and antigenome. Despite the critical role of RNA secondary structure in viral replication, pathogenesis, and immune evasion, there remains a striking paucity of research, both computational and experimental, dedicated to elucidating these features in RSV. By leveraging the synergistic power of these computational approaches, we identified eleven conserved and potential structural elements in the antigenome/mRNA strand and one novel structure in the genomic/−RNA strand, all of which exhibit characteristics suggestive of functional importance. These structures are hypothesized to play roles in processes such as transcription, replication, or interactions with host factors, offering potential targets for therapeutic intervention. Our findings fill a critical knowledge gap in RNA secondary structure in RSV biology, providing a foundational framework for future experimental validation. These observations mirror findings in other negative sense RNA viruses, such as Influenza A virus, where structured elements in coding regions regulate polymerase activity and genome packaging [8]. By combining ScanFold-driven thermodynamics, covariation analysis, ti/tv profiling and genomic localization, this study establishes a framework for prioritizing functional RNA structure in viral genomes. The multi-tiered approach mitigates limitations inherent to individual methods (e.g. ScanFold’s propensity to predict false-positive stable folds) by integrating evolutionary and mutational signatures, we were able to predict structure of functional importance. The eleven high-confidence structure identified here represent promising targets for broad-spectrum antivirals. For instance, small molecules disrupting the L gene’s conserved structures could inhibit polymerase activity, akin to ribavirin’s interference with HCV IRES elements. Similarly, antisense oligonucleotides (ASOs) targeting the F gene’s fusogenic structure might block conformational rearrangements required for membrane fusion. While this computational pipeline provides robust predictions, limitations of this study is the absence of experimental data to further buttress the identified structures. Experimental techniques like SHAPE or DMS probing could confirm in vivo and/or in vitro structural conformations. Another consideration is the restricted window size of the ScanFold program (120 nt), which inherently captures only localized snapshots of the functional RNA structurome of RSV. Despite this, similarly sized structures comprise > 90% of known RNA structural families [52] and play important roles in other RNA viruses. Additionally, ScanFold’s focus on these local and robustly predicted elements have shown strong correlations with biochemical probing experiments [15, 53], highlighting their robust nature. This finding is further strengthened by our use of two orthogonal measures of structural conservation on RSV evolution, which highlight the likely functionality of the regions identified by ScanFold. While we can’t speak about the precise functions of these structures, beyond interesting hypotheses raised by their ability to modulate interactions with predicted RBPs, our results provide a roadmap for future studies to address these interesting questions. For example, using reverse genetics systems and mutagenesis to assess the functional impact of structural disruptions. Additionally, further experimental work using biochemical structure probing can help to place these identified structures into their larger sequence contexts, as well as address the potential for RNA conformational dynamics in regulating RSV biology. Supplementary Material