I6mA-DNC: Prediction of DNA N6-Methyladenosine sites in rice genome based on dinucleotide representation using deep learning

Abstract

DNA methylation is a crucial epigenetic process. DNA N6-methyladenine is closely related to a variety of biological processes such as DNA replication, transcription, repair and cellular defense. In genome, N6-methyladenine (6 ​mA) sites are not uniformly distributed; therefore, it is required to determine the genomic locations of 6 ​mA for better comprehension of its biological functions. Although various experimental procedures have been used to identify 6 ​mA sites and yielded positive results, these biochemical techniques are expensive and time-consuming. In order to solve this problem and provide ease for future researches, it is indispensable to develop a robust and accurate computational model to find N6-methyladenine sites. With this regard, we introduce a deep learning-based computational model called i6mA-DNC to detect the N6-methyladenine sites in the rice genome. We split the DNA sequences into dinucleotide components and feed them to the model. This model automatically extracts optimal features from the pre-processed data using convolution neural network (CNN). Our proposed model i6mA-DNC obtained 89.20% of specificity, 88.01% of sensitivity, 88.60% of accuracy, and 0.772 of MCC. These results prove that our intelligent model achieved better success rates in all evaluation metrics than existing methods. Our model i6mA-DNC is expected to become a useful tool for academic research on N6-methyladenine sites identification. A user-friendly webserver has been established and made freely accessible at https://home.jbnu.ac.kr/NSCL/i6mA-DNC.htm.