Decoding circRNA translation: challenges and advances in computational method development

Abstract

In recent years, numerous studies have demonstrated that circRNAs play crucial biological roles through their capacity to encode functional proteins. Computational methods have become essential for investigating circRNA translation. In this review, we first outline circRNA biogenesis and translation mechanisms to establish the rationale for developing specialized computational strategies. We then summarize experimental techniques and existing databases that support computational method development. Subsequently, we provide a systematic introduction to existing circRNA translation analysis tools and their underlying algorithms, with emphasis on benchmarking the performance of sequence-based methods using a unified dataset. Our benchmarking revealed that: (1) cirCodAn achieved superior predictive accuracy while maintaining user accessibility; (2) the training data selection during method development critically impacts model performance. This review serves as a comprehensive reference for the selection and application of circRNA translation analysis methods and provides foundational guidance for the development and refinement of future computational tools.

Keywords: bioinformatics; circRNA; coding potential; function; translation.

FIGURE 1

Experimental techniques, translation mechanisms, databases, and computational strategies constitute the core components of circRNA translation research. These elements interact synergistically to advance the field. Experimental methods uncover translation mechanisms and validate circRNA-derived proteins, generating reliable data. Mechanism provide the theoretical foundation for computational strategy development. Both experimental and computational data are deposited into databases, which serve as resources for researchers and support further method development. In turn, computational analyses guide experimental validation and contribute newly predicted data to the databases.

FIGURE 2

circRNA biogenesis and translation mechanisms. (A) circRNA biogenesis occurs in the nucleus and can be promoted through four major pathways: (i) intron pairing-driven circularization, (ii) RNA-binding protein (RBP)-mediated circularization, (iii) exon skipping-driven circularization, and (iv) lariat-driven circularization. (B) circRNA translation occurs in the cytoplasm and proceeds via cap-independent mechanisms, including: (i) IRES-mediated initiation, (ii) m⁶A-mediated initiation, and (iii) EJC-mediated initiation.

FIGURE 3

Framework for circRNA translation analysis based on Ribo-seq data. During translation, ribosomes bind to RNA sequences and associated regions from RNase digestion. After RNase treatment, the ribosome-protected fragments are isolated, used to construct sequencing libraries, and then sequenced. The resulting reads are mapped to the reference genome, and unmapped reads may originate from circRNAs. These candidate fragments are further screened to identify potential translation events derived from circRNAs. Finally, various machine learning methods can be applied to assess the translational potential of circRNAs in greater detail.

FIGURE 4

Framework for circRNA translation analysis based on MS. MS directly detects protein fragments (peptides), which are then aligned to reference protein sequences to determine their origin. For circRNAs, potential circRNA-encoded proteins are first predicted, and the presence of these proteins is validated by matching the detected fragments to the predicted circRNA-derived sequences.

FIGURE 5

The workflow of linear RNA-based prediction tools adapted for circRNAs, including sequence preprocessing, feature engineering, primary network, and cORF screening.

FIGURE 6

Performance evaluation of circRNA translation analysis methods.