TL;DR
Researchers have updated and expanded CSCD, a database for cancer-specific circular RNAs (circRNAs), into CSCD2. This new version boasts a significantly larger dataset of samples and identified circRNAs, along with enhanced features for predicting and visualizing interactions between circRNAs, microRNAs, and RNA-binding proteins, making it a more powerful resource for cancer research.
What Problem Does This Paper Address?
Circular RNAs (circRNAs) are a fascinating class of RNA molecules that have gained significant attention for their roles in various biological processes, particularly in cancer. Unlike linear RNAs, circRNAs form a closed loop, which confers unique properties like resistance to degradation. This stability makes them promising candidates for diagnostic markers and suggests they play crucial regulatory roles. However, understanding these roles requires comprehensive data on circRNAs, especially those specific to cancer, and their interactions with other molecules like microRNAs (miRNAs) and RNA-binding proteins (RBPs). The previous CSCD database provided a valuable resource, but the rapid growth of sequencing data and the need for more detailed interactional information necessitated an update. CSCD2 aims to address this by providing a more extensive and integrated platform for exploring cancer-specific circRNAs and their regulatory networks.
Methods
The CSCD2 database was built using a multi-step process:
- Sample Collection: RNA-Seq data from a substantial number of human cancer and normal tissue/cell line samples were collected from public databases (ENCODE, SRA).
- CircRNA Identification: Reads from RNA-Seq data were mapped to the human genome. CircRNAs were then identified using a combination of four established algorithms: CIRI2, circRNA_finder, find_circ, and circexplorer2. To ensure consistency, a 2 bp mismatch tolerance was applied when merging results from different algorithms.
- CircRNA Expression Normalization: The SRPTM method was used to normalize the expression levels of identified circRNAs.
- MicroRNA-CircRNA Interaction Prediction: Potential interactions were predicted by scanning a 100 bp window around the back-splice site of each circRNA for microRNA response elements (MREs) using TargetScan and miRanda. Expression data for miRNAs from TCGA were also incorporated.
- RBP-CircRNA Interaction Prediction: Protein binding sites of RBPs were identified by integrating CLIP-Seq data from STARBASE and POSTAR2. Additional IHC staining and mass spectrometry data from HPA and dbDEPC were used to characterize RBP expression across cancer types.
- Full-Length and ORF Prediction: The R package FcircSEC was used to extract full-length circRNA sequences, and the getORF tool from EMBOSS was employed to predict open reading frames (ORFs), indicating potential translational capacity.
- Database Implementation: All collected data and predicted interactions were organized into interactive MySQL tables and presented through a user-friendly web interface built with ThinkPHP and JavaScript.
Datasets
- RNA-Seq Data: 1113 samples were collected, comprising 472 cancer tissue samples, 353 normal tissue samples, 91 cancer cell line samples, and 197 normal cell line samples. These were sourced from the ENCODE and SRA databases.
- Gene Annotation: GENCODE (version 32) and Refseq (version GCF_000001405.39) were used for gene annotation.
- miRNA Expression Data: TCGA database was used to extract miRNA expression in 32 types of cancer.
- RBP Binding Data: CLIP-Seq data from STARBASE and POSTAR2 were integrated, covering 207 RBPs.
- RBP Expression Data: IHC staining and mass spectrometry data from HPA and dbDEPC were utilized.
Key Insights and Ideas
- The sheer scale of circRNA diversity: CSCD2 identifies nearly 3 million circRNAs, highlighting the vastness of this RNA landscape, with over a million identified as cancer-specific.
- Beyond identification: Understanding function: The integration of miRNA and RBP interaction predictions, along with full-length and ORF sequences, moves beyond simply cataloging circRNAs to providing tools for understanding their regulatory mechanisms and potential translational roles.
- CircRNAs as potential biomarkers: The focus on cancer-specific circRNAs underscores their potential utility in cancer diagnosis and prognosis, given their unique expression patterns in disease states.
- The importance of comprehensive data integration: By combining data from multiple algorithms, public databases, and prediction tools, CSCD2 offers a more robust and reliable resource than single-source databases.
Key Results
- Expanded circRNA repertoire: CSCD2 contains 2,901,587 identified circRNAs, a significant increase from previous resources. This includes 1,013,461 cancer-specific circRNAs (CS-circRNAs), 1,533,704 from normal samples, and 354,422 common to both.
- Rich interaction data: The database predicts over 175 million miRNA-circRNA interactions and over 1.375 billion RBP-circRNA interactions.
- Functional potential indicators: Predictions for full-length sequences and ORFs are provided for a substantial number of circRNAs, offering insights into their translational capabilities.
- User-friendly interface: CSCD2 offers interactive web interfaces for browsing and searching circRNAs, miRNAs, and RBPs, facilitating data exploration and analysis.
Limitations and Future Work
- Computational predictions: While extensive, the miRNA-circRNA and RBP-circRNA interactions are based on computational predictions (binding motifs, CLIP-Seq data). Experimental validation would be necessary to confirm these interactions.
- Debate on functional significance: The paper acknowledges ongoing debate regarding the functional significance of some circRNAs, suggesting that not all identified circRNAs may have a direct biological role.
- Future updates: The authors plan to regularly update the database with new data and functions, especially with the advancement of technologies like nanopore sequencing for circRNA research. This suggests a commitment to keeping CSCD2 at the forefront of circRNA research.
Summary updated