Mouse4mC-BGRU: Deep learning for predicting DNA N4-methylcytosine sites in mouse genome.

Abstract

DNA N4-methylcytosine (4mC) is an important DNA modification and plays a crucial role in a variety of biological processes. Accurate 4mC site identification is fundamental to improving the understanding of 4mC biological functions and mechanisms. However, lots of identification approaches are limited to traditional machine learning, which leads to weak learning ability and a complex feature extraction process. Here, we propose Mouse4mC-BGRU, an advanced deep learning model that utilizes adaptive embedding based on bidirectional gated recurrent units (BGRU). Benchmark results show that our model performs better than the state-of-the-art methods in the prediction of 4mC sites in the mouse genome. By using adaptive features to extract representation, Mouse4mC-BGRU can capture the latent biology information of input sequence, which effectively enhances model representation ability. In addition, we visualize the training process of Mouse4mC-BGRU with dim reduction tools and intuitively show the effectiveness of our model, demonstrating that Mouse4mC-BGRU has great potential to be a powerful and practically useful tool to accurately identify 4mC sites.