Application of Deep Learning in the Identification of Extrachromosomal Circular DNA (eccDNA)

EccDNA has emerged as a pivotal factor in regulating oncogene expression. Its diverse structures and rich variability present significant challenges for accurate identification. In this study, we propose a composite deep learning model that integrates convolutional neural networks (CNNs), residual networks, and attention mechanisms. Leveraging existing eccDNA datasets, the model is designed to automatically detect eccDNAs across various samples, with the goal of enhancing recognition efficiency and precision. Upon training, the model demonstrated a significant improvement in eccDNA identification. Additionally, we developed a software tool named ECCNET to facilitate access and application by researchers. Validation experiments confirm that the model exhibits outstanding performance in eccDNA sequence recognition and analysis, maintaining a high degree of consistency across different datasets. This research effectively applies advanced deep learning techniques to eccDNA identification, laying a solid foundation for further exploration in this field.