Prediction of circRNA-miRNA interaction sites based on fully convolutional neural network and gated recurrent unit
Computational Biomedicine · 2026
DOI: 10.70401/cbm.2026.0025 ↗Kurz gefasst
Fachartikel in der Zeitschrift Computational Biomedicine, 2026. Verfasst von Ya Qiu und 4 weiteren Autor:innen.
Abstract (Originalsprache)
Aims
Circular RNAs (circRNAs) act as vital microRNA (miRNA) sponges and participate in the pathogenesis of various complex diseases. Most existing computational methods focus on sequence-level prediction of circRNA-miRNA interactions, while few tools can accurately locate interaction sites at the nucleotide level.
Methods
In this study, we propose CMIFCG, a novel model combining fully convolutional neural networks (FCNs) and bidirectional gated recurrent units (BiGRU) for nucleotide-level prediction of circRNA-miRNA interaction sites. We adopted one-hot encoding for sequence preprocessing, utilized stacked convolutional modules to extract high-order features, and integrated BiGRU to capture sequential association features. The decoding module with skip connections restores feature maps to the original sequence size for precise site prediction. We constructed four datasets with different sequence lengths and negative samples to evaluate model performance.
Results
Experimental results show that CMIFCG achieves better performance than other baseline models, and excessive module stacking causes performance degradation. Motif analysis confirms that the predicted motifs match well with known miRNA sequences. Partial predicted interactions are further verified via public databases.
Conclusion
In summary, the CMIFCG model enables effective fine-grained identification of circRNA-miRNA interaction sites, providing a new tool for exploring circRNA regulatory mechanisms and disease-related molecular pathways.
Zitieren
APA
Qiu, Y., Li, Y., Zhao, Y., Zhou, P., & Shen, Z. (2026). Prediction of circRNA-miRNA interaction sites based on fully convolutional neural network and gated recurrent unit. Computational Biomedicine. https://doi.org/10.70401/cbm.2026.0025
Vancouver
Qiu Y, Li Y, Zhao Y, Zhou P, Shen Z. Prediction of circRNA-miRNA interaction sites based on fully convolutional neural network and gated recurrent unit. Computational Biomedicine. 2026. doi:10.70401/cbm.2026.0025
BibTeX
@article{Qiu2026,
author = {Qiu, Ya and Li, Yulu and Zhao, Yike and Zhou, Pengyu and Shen, Zhen},
title = {Prediction of circRNA-miRNA interaction sites based on fully convolutional neural network and gated recurrent unit},
journal = {Computational Biomedicine},
year = {2026},
doi = {10.70401/cbm.2026.0025}
}
EUVIMED