SpineNet-6mA: A Novel Deep Learning Tool for Predicting DNA N6-Methyladenine Sites in Genomes

Abstract

DNA N6-methyladenine (6mA) has subsequently been identified as an important epigenetic modification which plays an important role in various cellular processes. The precise discrimination of N6-methyladenine (6mA) in genomes is required to recognize its biological functions. Although, we have several experimental techniques for the identification of 6mA-sites, in silico prediction has evolved as an alternative approach due to high-cost and labor-intense in experimental techniques. Taking into account, the implementation of an efficient and accurate model for identification of N6-methyladenine is of high priority. Several machine learning and deep learning models have already been developed to classify genome-wide 6mA sites. However, their success in predicting 6mA sites still has room for improvement. Based on this, we proposed a novel deep learning based model for the prediction of DNA N6-methyladenine sites in rice genomes. We built our model based on a special architecture called SpinalNet using DNA 6mA sites in rice genome and obtained an accuracies of 94.31% and 94.77% with an MCCs of 0.88 and 0.89 on two different datasets. The model generalizes well to other genomes as well, validated through cross-species testing. The results validate that the proposed model produces better scores than existing models regarding all evaluation parameters. A user-friendly webserver is made available at http://nsclbio.jbnu.ac.kr/tools/SpineNet6mA/.

Keywords

• Deep learning
• DNA sequence
• epigenetics
• neural networks
• spinalnet