RPS 2.0: an updated database of RNAs involved in liquid–liquid phase separation

Abstract

Liquid–liquid phase separation (LLPS) is a crucial process for the formation of biomolecular condensates such as coacervate droplets, P-bodies and stress granules, which play critical roles in many physiological and pathological processes. Increasing studies have shown that not only proteins but also RNAs play a critical role in LLPS. To host LLPS-associated RNAs, we previously developed a database named ‘RPS’ in 2021. In this study, we present an updated version RPS 2.0 (https://rps.renlab.cn/) to incorporate the newly generated data and to host new LLPS-associated RNAs driven by post-transcriptional regulatory mechanisms. Currently, RPS 2.0 hosts 171 301 entries of LLPS-associated RNAs in 24 different biomolecular condensates with four evidence types, including ‘Reviewed’, ‘High-throughput (LLPS enrichment)’, ‘High-throughput (LLPS perturbation)’ and ‘Predicted’, and five event types, including ‘Expression’, ‘APA’, ‘AS’, ‘A-to-I’ and ‘Modification’. Additionally, extensive annotations of LLPS-associated RNAs are provided in RPS 2.0, including RNA sequence and structure features, RNA–protein/RNA–RNA interactions, RNA modifications, as well as diseases related annotations. We expect that RPS 2.0 will further promote research of LLPS-associated RNAs and deepen our understanding of the biological functions and regulatory mechanisms of LLPS.

Graphical Abstract

Graphical Abstract.

Introduction

Biomolecular condensates, such as nucleoli, P-bodies and stress granules (SGs), are dynamic, membraneless assemblies of proteins, RNAs and other biomolecules that form by liquid-liquid phase separation (LLPS). Accumulating studies have demonstrated that LLPS-mediated condensates play critical roles in the regulation of numerous biological processes, including cellular stress responses (1), chromatin organization (2), gene transcription (3), RNA splicing (4) and protein degradation (5). The dysregulation of LLPS is closely linked to various diseases, such as cancer (6), neurodegenerative diseases (7), viral infection (8) and autoimmune diseases (9).

An increasing number of studies have highlighted the fundamental roles of RNAs in LLPS. Many RNAs act as scaffolds for LLPS (10) or play key roles through multivalent interactions with RNA-binding domains of proteins (11). While previous research on LLPS primarily focused on the protein components, leading to the development of several databases focused on LLPS-associated proteins, such as LLPSDB (12), PhaSePro (13), PhaSepDB (14), DrLLPS (15), RNAgranuleDB (16) and HUMANCELLMAP (17), there remains a significant lack of resources specifically for LLPS-associated RNAs. To fill this gap, we previously developed a database called ‘RPS’ (18), which cataloged 21 613 LLPS-associated RNAs across 20 distinct biomolecular condensates and has been widely used (19,20). As LLPS research continues to evolve, an accumulating body of studies focusing on LLPS-associated RNAs have emerged. Moreover, growing attention has been directed to explore the impact of post-transcriptional regulatory events on LLPS, such as alternative polyadenylation (APA) (21,22), alternative splicing (AS) (4,23), adenosine-to-inosine (A-to-I) RNA editing (24), and RNA modifications (25–27). For example, A-to-I editing of lncRNA such as NEAT1, NORAD and XIST can alter their structures and RNA-RBP interaction sites to regulate LLPS (24). m6A modifications on RNA can promote LLPS through YTHDF2-mediated interactions (25). These data provide a new perspective for the identification of LLPS-associated RNAs. Therefore, there is an urgent need to update RPS (18) to integrate these expanding data and to host new LLPS-associated RNAs driven by post-transcriptional regulatory mechanisms.

In this study, we present an updated database, RPS 2.0 (https://rps.renlab.cn/). Currently, RPS 2.0 contains 171 301 entries of LLPS-associated RNAs across 24 different biomolecular condensates which are categorized into four evidence types: ‘Reviewed’, ‘High-throughput (LLPS enrichment)’, ‘High-throughput (LLPS perturbation)’ and ‘Predicted’ and five event types (Figure 1). Notably, RPS 2.0 integrates post-transcriptional regulatory events associated with LLPS. To facilitate further investigation into the functions of LLPS-associated RNAs and their relationship to diseases, RPS 2.0 not only provides basic information on LLPS-associated RNAs along with experimental and high-throughput evidence supporting these associations, but also provides abundant RNA annotations (sequences, RNA secondary structures, RNA–RNA/RBP binding sites and RNA modifications) and disease annotations. With these significant upgrades, we expect that RPS 2.0 will serve as a valuable resource for research on LLPS-associated RNAs.

Figure 1.

Overall design and construction of RPS 2.0. LLPS-associated RNAs deposited in RPS 2.0 were derived from literature mining, high-throughput sequencing experiments, and predictions based on interaction networks. RPS 2.0 offers detailed information on LLPS-associated RNAs and LLPS processes. In addition, advanced annotations of LLPS-associated RNAs are provided based on multiple databases and bioinformatics tools. This figure is created with BioGDP.com (https://BioGDP.com).

Materials and methods

Data collection

All literature and high-throughput datasets in RPS 2.0 were collected from NCBI PubMed, GEO (28), EMBL-EBI (29) and DDBJ (30) by the following keyword combinations: ‘((phase transition[Title/Abstract]) OR (phase separation[Title/Abstract]) OR (biomolecular condensates[Title/Abstract]) OR (membraneless organelle[Title/Abstract]) OR (G body[Title/Abstract]) OR (stress granule[Title/Abstract]) OR (P-body[Title/Abstract]) OR (midbody[Title/Abstract])) AND (RNA[Title/Abstract])’. Based on the data sources, RPS 2.0 categorized LLPS-associated RNAs into four evidence types: (i) Reviewed: We initially retrieved a total of 3131 papers published up to 30 June 2024. After manually reviewing the abstracts or full articles, we finally screened out 262 publications that provided the experimental evidence of LLPS-associated RNAs (Supplementary Table S1). Both natural and synthetic, full-length and truncated RNAs were curated. (ii) High-throughput (LLPS enrichment): LLPS-associated RNAs from the ‘High-throughput (LLPS enrichment)’ level were identified from high-throughput datasets of purified condensates. A total of 12 datasets including 140 samples from humans, mice, yeast, and Chinese hamsters were collected (Supplementary Table S2). Subsequently, we analyzed these datasets to identify RNAs enriched in LLPS condensates by comparing the RNA content within the condensate to that in the cytoplasm. (iii) High-throughput (LLPS perturbation): LLPS-associated RNAs from the ‘High-throughput (LLPS perturbation)’ level were identified from high-throughput datasets generated from experiments that perturb key LLPS molecules or disrupt phase separation structures. A total of eight datasets including 88 samples from humans, mice, yeast, and worms were collected (Supplementary Table S2). We then analyzed these datasets to identify RNAs affected after LLPS perturbation by comparing the perturbation group to the control group. (iv) Predicted: we identified RNAs that are likely to undergo LLPS through RNA–protein interactions. LLPS-associated proteins were obtained from six databases, including LLPSDB (12), PhaSePro (13), PhaSepDB (14), DrLLPS (15), RNAgranuleDB (16) and Human cell Map (17). We then filtered out proteins associated only with RNA-free condensates (such as receptor cluster and Z granule), retaining 9 581 LLPS-associated proteins for prediction. Based on experimentally validated RNA-RBP interactions obtained from NPInter (31) and RNAInter (32), 5795 RNA interactors with these retained LLPS-associated proteins were identified.

Curation of LLPS processes

Detailed information was curated for all LLPS processes, including LLPS observation methods such as fluorescence recovery after photobleaching (FRAP) and dynamic light scattering (DLS), as well as RNA detection methods such as smFISH, immunostaining, and RNA-seq. Additionally, experimental conditions like pH, temperature, RNA concentration, and RNA/protein ratios were documented. However, ‘Predicted’ LLPS-associated RNAs do not have corresponding information on specific LLPS processes.

Comprehensive analysis of high-throughput datasets

All high-throughput datasets collected were uniformly processed following the standardized procedure outlined below. Initially, raw reads were processed using fastp (33) to filter out low-quality reads and remove adapters. The trimmed reads were then aligned to the reference genomes (Homo sapiens: hg38, Mus musculus: mm39, Saccharomyces cerevisiae: SacCer3, Caenorhabditis elegans: ce11, Cricetulus griseus: criGriChoV2) using STAR (34). For LLPS-associated expression events, the aligned reads were quantified using featureCounts (35). Samples within each dataset were classified into the condensate group and the control group (e.g. RNA granule versus cytoplasm) or into the perturbation group and the control group (e.g. NEAT1 KO vs. no treatment). DESeq2 (36) was employed to identify the differentially expressed genes (log₂FC > 1 for RNAs of ‘LLPS enrichment’, |log₂FC| > 1 for RNAs of ‘LLPS perturbation’, P-value < 0.05). For LLPS-associated APA events, DaPars (37) was applied to analyze the differential APA levels between two groups, with parameters set to Coverage_cutoff = 30, FDR_cutoff = 0.05, PDUI_cutoff = 0.5, Fold_change_cutoff = 0.26 (equivalent to 1.2-fold change) and a significance threshold of P < 0.05. To analyze LLPS-associated AS events, rMATS (38) was utilized to identify differential AS between two groups with default parameters and a significance threshold of P < 0.05. For LLPS-associated A-to-I events, REDItools (39) was employed to identify A-to-I sites. The results were filtered to include only A-to-I events with a frequency less than 0.95 and coverage greater than 10, followed by a Wilcoxon test with a significance threshold of P < 0.1. Finally, to investigate LLPS-associated modification events, ModTect (40) was used to detect potential RNA modification sites and comparisons between paired groups were performed using a Wilcoxon test with a significance threshold of P < 0.1.

Annotation of LLPS-related RNAs

RPS 2.0 provided fundamental information of LLPS-associated RNAs, encompassing evidence type, event type, genomic location, gene symbol, gene ID, transcript ID, RNA type, sequence, and functional roles in LLPS. These data were preferentially extracted from the original publications, otherwise, they were acquired from Ensembl (41) or UCSC (42) databases. For reviewed LLPS-associated RNAs, sequences were primarily extracted from the original publications. Sequences for other RNAs were retrieved from databases such as Gencode (43), Ensembl (41), NCBI (44) and miRbase (45), and are represented by canonical transcripts.

RPS 2.0 also integrated multiple resources and tools to annotate distinct LLPS-associated properties of these RNAs. Sequence-specific features, including tandem repeats and RNA sequence families, were detected separately using tandem repeat finder (TRF) (46) and Rfam web server (47). For structure annotation, we first extracted information on LLPS-associated structures from the original publications. Then, we predicted RNA secondary structures based on a minimum free energy (MFE) algorithm using RNAfold from ViennaRNA package (48). G-quadruplex structures were predicted using the G4RNA screener web server (49) with the default settings. Besides, we identified known sequence and secondary structure protein binding motifs in RNAs from humans and mice via BRIO web server (50). Moreover, the RNA-protein and RNA-RNA binding sites were acquired from ENCORI (51), POSTAR3 (52), NPInter v5 (31), and RNAInter v4 (32). The RNA modification sites are integrated from RMBase v3 (53), m5C-Atlas (54) and m7GHub v2 (55).

To discover potential relations between LLPS-associated RNAs and diseases, we collected RNA-disease associations with experimental evidence from RNADisease v4 (56), LncRNADisease v3 (57) and DisGeNET (58). Furthermore, we also integrated disease-associated variants from DisGeNET (58) and ClinVar (59), along with cancer-associated variants taken from COSMIC (60), ICGC (61) and TCGA (62). The genomic coordinates of all data resources were further converted to the reference genome versions used in RPS 2.0 using the LiftOver program (42).

Database and web interface implementation

All data in RPS 2.0 were stored and managed using MySQL tables. The server-backend was developed based on Java, while the web-frontend interfaces were implemented using HyperText Markup Language (HTML), Cascading Style Sheets (CSS) and JavaScript (JS). All the interactive charts were generated by ECharts to visualize the analysis results. Furthermore, RPS 2.0 implemented a genome browser to present genomic annotations using UCSC Genome Browser (http://genome.ucsc.edu/) (42).

Results

Significant enhancements in data content of RPS 2.0

Given the essential roles of RNAs in liquid-liquid phase separation (LLPS), we and other researchers initially developed RPS 1.0 (18) and RNAPhaSep (63) to host LLPS-associated RNAs. Compared to RPS 1.0 (18) and RNAPhaSep (63), RPS 2.0 is far superior in all aspects (Table 1). Specifically, RNAPhaSep (63) focused exclusively on LLPS-associated RNAs supported by validated literature, covering 166 references. Meanwhile, RPS 1.0 (18) included 21 613 LLPS-associated RNAs across 20 distinct biomolecular condensates derived from 131 published studies and 7 high-throughput datasets. RPS 2.0 has significantly expanded its data. It now encompasses 171 301 LLPS-associated RNAs spanning 15 organisms and 24 condensates, sourced from 262 published studies and 20 high-throughput datasets, which is more than twice the number of RPS 1.0 (18). And it features 61 211 protein-coding RNAs (up from 19 655 in RPS 1.0), 10 166 lncRNAs (up from 3632 in RPS 1.0), 3887 pseudogene RNAs (up from 1387 in RPS 1.0), and 3407 other RNAs (up from 1855 in RPS 1.0). In comparison, RNAPhaSep (63) includes only 123 protein-coding RNAs, 78 lncRNAs, and 229 other RNAs (Table 1). In addition to integrating the latest literature and related high-throughput datasets of purified condensates, RPS 2.0 now includes high-throughput datasets from experiments that perturb key molecules of LLPS or disrupt phase separation structures, which are absent in either RPS 1.0 (18) or RNAPhaSep (63) (Table 1). According to the data source, we have categorized the LLPS-associated RNAs in RPS 2.0 into four evidence levels: (i) Reviewed: This category includes 517 LLPS-associated RNAs that have been experimentally validated to participate in LLPS through in vivo or in vitro experiments from 262 published studies. Among these, 184 RNAs are synthetic, while 333 are natural. (ii) High-throughput (LLPS enrichment): this category contains 112 277 LLPS-associated RNAs that were enriched in purified condensates compared to cytoplasmic total RNAs, based on analyses of high-throughput sequencing data of six condensates across 32 LLPS processes. For example, the expression level of AHNAK mRNA has been shown to be significantly up-regulated in stress granules compared to cytoplasmic total RNAs (Supplementary Figure S1A), indicating its enrichment and critical role in stress granule formation. (iii) High-throughput (LLPS perturbation): this includes 54 375 LLPS-associated RNAs that were altered upon perturbation of key LLPS molecules or disruption of phase separation structures. This analysis was based on high-throughput sequencing data of four condensates across 20 LLPS processes. For instance, NaN₃ treatment is known to promote the formation of stress granules (64). High-throughput data analysis has revealed that the expression level of NIP1 mRNA was significantly down-regulated after NaN₃ treatment (Supplementary Figure S1B). Since NIP1 is a subunit of eukaryotic translation initiation factor 3 (eIF3) and plays an important role in protein translation (65), this down-regulation could indicate that protein translation is paused due to the formation of stress granules under oxidative stress. (iv) Predicted: This category comprises 5795 RNAs that were found to interact with screened LLPS-associated proteins. Notably, we observed that 85.65% of the ‘Reviewed’ LLPS-associated RNAs were validated by other evidence types, underscoring the reliability of the LLPS-associated RNAs included in RPS 2.0 (Figure 2A).

Table 1.

Comparison among databases related to LLPS-associated RNAs

Content of database	RPS 2.0	RPS 1.0	RNAPhaSep
Literature	262	131	166
Organism	15	13	22
Condensates	24	20	15
High throughput datasets	20	7	×
Number of protein-coding RNA	61 211	19 655	123
Number of lncRNA	10 166	3632	78
Number of pseudogene RNA	3887	1387	0
Number of other RNA	3407	1855	229
LLPS perturbation datasets	√	×	×
Post-transcriptional events	√	×	×
RNA structure	√	√	√
G-quadruplex structures	√	√	×
Protein binding structure motifs	√	√	×
Protein binding sequence motifs	√	√	×
RNA families	√	√	×
Tandem repeats	√	√	×
RNA–RBP binding sites	√	√	×
RNA–RNA binding sites	√	√	×
Modifications	√	√	×
Disease annotations	√	√	√

Figure 2.

Characteristics of LLPS-associated RNAs in RPS 2.0. (A) The overlap between LLPS-associated RNAs with different evidence types in the shared organisms, regardless of the condensate types. (B) Comparison of sequence lengths between LLPS-associated and non-LLPS-associated RNAs. Wilcoxon signed-rank test, **** P < 0.0001. (C) Comparison of GC content between LLPS-associated and non-LLPS-associated RNAs. Kolmogorov–Smirnov test. (D) The proportion of RNAs with tandem repeats in LLPS-associated and non-LLPS-associated RNAs (left). Two-proportion z-test, **** P < 0.0001. Period size distribution of tandem repeats in LLPS-associated and non-LLPS-associated RNAs (right). (E) The proportion of RNAs with G4 structures in LLPS-associated and non-LLPS-associated RNAs. Two-proportion z-test, **** P < 0.0001. (F) Comparison of modification sites between LLPS-associated and non-LLPS-associated RNAs. Wilcoxon signed-rank test, **** P < 0.0001. (G) Comparison of the number of interactors (RBP and RNA) between LLPS-associated RNAs and non-LLPS-associated RNAs. Wilcoxon signed-rank test, **** P-value < 0.0001.

Another significant enhancement in RPS 2.0 is the incorporation of LLPS-associated post-transcriptional regulatory events, which broadens our understanding of the regulatory mechanisms involved in phase separation. Specifically, RPS 2.0 includes a total of 3931 RNA modification events, 2780 A-to-I editing events, 41 506 alternative polyadenylation (APA) events and 40 549 alternative splicing (AS) events (Table 2). These diverse and extensive post-transcriptional regulatory events highlight their crucial roles in both the formation and regulation of LLPS.

Table 2.

LLPS-associated RNAs in distinct condensates with different event types

Condensate	Expression	Modification	Alternative splicing	APA	A-to-I	Total
Stress granule	31 065	1812	19 020	24 487	1140	77 524
Germ plasm	13 762	53	1245	500	−	15560
P-body	7491	1432	6758	8288	811	24 780
Cytoplasmic granule	5654	472	3856	4529	409	14 920
Paraspeckle	5286	155	8438	3702	420	18 001
G body	334	−	−	−	−	334
Coacervate droplet	252	7	5	−	−	264
Midbody	70	−	1227	−	−	1297
Nuclear body	12	−	−	−	−	12
Nuclear speckle	12	−	−	−	−	12
Nucleolus	6	−	−	−	−	6
Others	10	−	−	−	−	10

Characteristics of LLPS-associated RNAs

To further characterize LLPS-associated RNAs, we performed a comparative analysis of various RNA properties between LLPS-associated RNAs and non-LLPS RNAs in RPS 2.0. Non-LLPS RNAs were defined as canonical transcripts that were not deposited in RPS 2.0. Consistent with previous studies (66,67), we found that various kind of RNAs involved in LLPS are significantly longer than the corresponding non-LLPS RNAs (Figure 2B, P-value < 0.0001, Wilcoxon signed-rank test), suggesting that longer RNAs may function as extended scaffolds, providing more interaction sites necessary for LLPS. We also found that the GC content of LLPS-associated RNAs is slightly higher (Figure 2C, P-value < 0.0001, Kolmogorov-Smirnov test). However, previous studies have reported that GC content is not significantly correlated with stress granule enrichment (68,69) and is generally lower in RNAs enriched in P-bodies (70). These contrasting observations suggest that the influence of GC content may vary depending on the type of cellular condensate (71). Additionally, the proportion of RNAs with tandem repeats is higher in LLPS-associated RNAs than in non-LLPS RNAs (Figure 2D, P-value < 0.0001, two-proportions z-test). This is consistent with previous findings that low-complexity repeat sequences can drive LLPS when the number of repeat units exceeds a critical threshold (72). Our data suggest that RNAs with repeat numbers between 10 and 95 are most prone to undergo LLPS, while RNAs with too few or too many tandem repeats are less likely to participate in this process (Figure 2D). Moreover, we observed that LLPS-associated RNAs contain more G-quadruplex (G4) structures than non-LLPS RNAs (Figure 2E, P-value < 0.0001, two-proportions z-test), which are known to regulate LLPS in a sequence-specific manner (73). In addition, RNA modifications such as m6A, m5C, and m7G are more prevalent in LLPS-associated RNAs compared to non-LLPS RNAs (Figure 2F, P-value < 0.0001, Wilcoxon signed-rank test), indicating their potential role in modulating LLPS (74). We found that LLPS-associated RNAs have significantly more interaction sites than non-LLPS RNAs (Figure 2G, P-value < 0.0001, Wilcoxon signed-rank test). This suggests that increased multivalent interactions are likely to promote LLPS, which is in line with previous findings (75).

To investigate the details of the association between LLPS and post-transcriptional regulatory events, we conducted a detailed analysis using LLPS-associated RNAs from evidence type of ‘High-throughput (LLPS enrichment)’. We observed that each post-transcriptional event type has RNA sets distinct with RNA expression event, indicating that post-transcriptional events can occur without altering expression levels (Figure 3A, B). For ‘APA’ events, LLPS-associated RNAs are more likely to have extended 3′ UTRs (Figure 3C). In the case of ‘AS’ events, all splice types, particularly ‘Mutually Exclusive Exons (MXE)’ and ‘Retained Intron (RI)’ are more frequent outside of condensates (Figure 3D). Additionally, our analysis revealed that compared to non-LLPS RNAs, A-to-I editing sites in LLPS-associated RNAs are significantly enriched in the 3′ UTR but depleted in the intronic regions (Figure 3E). Regarding RNA modifications, we found that A, T and C bases are more susceptible to be modified within condensates, which with higher modification scores predicted by ModTect (40) (Figure 3F). To explore the potential biological functions of LLPS, we performed GO analysis on RNAs obtained from the ‘High-throughput (LLPS perturbation)’ evidence. As expected, the result shows significant enrichment in pathways associated with organelle localization and MLO assembly (Figure 3G). Moreover, pathways related to cell division and DNA repair were also found to be enriched, indicating critical roles of LLPS in these pathways (Figure 3G). Finally, the proportion of disease-related LLPS RNAs is significantly higher than that of non-LLPS RNAs (Figure 3H, P-value < 0.0001, two-proportions z-test), indicating that RNA-mediated phase separation is closely related to human diseases.

Figure 3.

Post-transcriptional regulatory events associated with LLPS and functional analysis of LLPS. (A) The set size and intersection numbers of LLPS-associated RNAs across different event types. (B) The percentage of unique versus overlapping RNA sets across different event types. (C) Change in PDUI (Percentage of Distal Polyadenylation Isoforms) value from the comparison between condensates and total RNAs. (D) The number of RNAs with different PSI (Percent Spliced In) value from the comparison between condensates and total RNAs. (E) The genomic location of enriched (outer circle) or depleted (inner circle) A-to-I sites in LLPS-associated RNAs compared to total RNAs. (F) Change in modification score across A, C, G and T bases from the comparison between condensates and total RNAs. (G) GO analysis of RNAs from ‘High-throughput (LLPS perturbation)’ evidence. (H) The proportion of disease-related RNAs in LLPS-associated RNAs and non-LLPS-associated RNAs. Two-proportion z-test, **** P < 0.0001.

Web interface and usage

RPS 2.0 provides a user-friendly web interface that enables users to quickly and interactively explore any LLPS-associated RNA or process (Figure 4). Compared to RPS 1.0 (18), the browse and detail pages have been improved to provide much clearer information on LLPS-associated RNAs.

Figure 4.

Basic functions of RPS 2.0 web interface. (A) The browsing interface of RPS 2.0. (B) The options available on the RPS 2.0 search page. (C) Detailed information and advanced annotations of LLPS-associated RNAs can be accessed by clicking on the corresponding RPS ID, while LLPS process information related to each RNA can be viewed by clicking on the corresponding LLPS ID. This figure is created with BioGDP.com (https://BioGDP.com).

Browse and search, on the browse page (Figure 4A), users can explore the LLPS-associated RNAs by choosing different organisms or condensates. After selecting a specific LLPS condensate, a detailed information card about the selected condensate is displayed, followed by a statistics card that includes charts depicting various evidence types and event types, along with diagrams illustrating the distribution of different RNA categories in different evidence types. A browse table is provided to show the query result. Compared to RPS 1.0 (18), the structure of the browse table has changed. We assign an ‘RPS ID’ to every entry by merging identical sequences from the same organism, condensate, and event type in RPS 2.0. In addition, to help users to filter out noises, we provide the number of LLPS processes supported by literature and high-throughput datasets, the predicted status and confidence score in the browse table. The confidence score is calculated based on the total number of LLPS processes supported by different evidence types. RNA entries are sorted based on their confidence scores. On the search page (Figure 4B), users can search LLPS-associated RNAs by various filters including gene symbol, gene ID, transcript ID, organism, natural/synthetic, RNA type, condensate, evidence type, RPS ID and event type.

Detail, by clicking on the RPS ID in the browse or search result table, users can access comprehensive information for specific LLPS-associated RNAs (Figure 4C). The details include basic information, RPS details, and advanced annotations. Compared to RPS 1.0 (18), the ‘RPS details’ section contains all details of supported evidence types including ‘Reviewed’, ‘High-throughput (LLPS enrichment)’, ‘High-throughput (LLPS perturbation)’, ‘Predicted’. Following this, annotations of various RNA properties, including sequence features, structures, RNA−protein/RNA interactions, RNA modifications, and annotations of disease associations are provided. By clicking the ‘LLPS ID’ on the RPS page, users can get detailed information about the specific LLPS process, such as experiment conditions, detection methods for LLPS and RNA, LLPS-associated diseases, corresponding PMID, further description of LLPS and all the RNA components involved in this LLPS process

RPS 2.0 allows users to download all LLPS-associated RNAs information of different evidence types from the ‘Download’ page. Detailed guidance on the usage of RPS 2.0 are available on the ‘Help’ page.

Summary and perspectives

RPS 2.0 is an updated database of RNAs involved in LLPS. It features the most extensive data content and an optimized structure among databases hosting LLPS-associated RNAs. In addition to integrating the latest literature and analyses of high-throughput datasets of purified condensates, RPS 2.0 now includes high-throughput datasets from experiments that perturb key molecules of LLPS or disrupt phase separation structures. This expansion enriches the resources available to researchers investigating the functions and mechanisms of LLPS.

Furthermore, RPS 2.0 covers plentiful LLPS-associated RNAs from five post-transcriptional regulatory events. These diverse and extensive post-transcriptional regulatory events highlight their crucial roles in both the formation and regulation of LLPS. RNA modifications can influence RNA stability, localization, and interaction with proteins, all of which are critical factors in the dynamics of biomolecular condensates. Similarly, A-to-I editing can alter the coding potential of RNAs, potentially affecting protein function and interactions within condensates. Furthermore, APA and AS are essential processes that contribute to the diversity of the transcriptome, allowing cells to fine-tune gene expression in response to various physiological conditions. By providing detailed insights into these post-transcriptional events, RPS 2.0 enables researchers to explore how these regulatory mechanisms might affect LLPS behavior and the functional outcomes of condensate formation.

We expect that RPS 2.0 will provide comprehensive support for the investigation of LLPS-associated RNAs, enabling users to efficiently explore their characteristics and regulatory mechanisms. We believe RPS 2.0 will empower researchers to develop therapeutic strategies for LLPS-related diseases and provide additional intervention approaches. Furthermore, we are committed to ensuring the long-term maintenance and reliability of the RPS database through continuous updates and rigorous validation.

Data availability

RPS 2.0 is a comprehensive online database available at https://rps.renlab.cn/.

Supplementary data

Supplementary Data are available at NAR Online.

Funding

National Key Research and Development Program of China [2023YFF1204600]; Guangdong Basic and Applied Basic Research Foundation [2021B1515020108]; National Natural Science Foundation of China [32470709]. Funding for open access charge: National Key Research and Development Program of China [2023YFF1204600].

Conflict of interest statement. None declared.