IN6-Methyl (5-step): Identifying RNA N6-methyladenosine sites using deep learning mode via Chou's 5-step rules and Chou's general PseKNC

Abstract

N6-methyladenosine (m6A) is an RNA methylation modification and it is involved in various biological progresses such as translation, alternative splicing, degradation, stability, etc. Therefore, it is highly recommended to develop computational models for detecting N6-methyldenosine sites in RNA as experimental technologies, such as m6A-seq and MeRIP-Seq, are both expensive and time consuming. Previous works start with features design step, which requires domain knowledge, followed by a classifier or cascade of classifiers for m6A sites identification. In this paper, on the other hand, we utilize an automatic feature learning approach based on the widely used natural language technique “word2vec”. The learnt features are extracted automatically from the human genome without any explicit definition. Then, these learnt features are fed to a simple convolution neural network model for classification. The proposed model is denoted as “iN6-Methyl (5-step)”. It has been evaluated on three publicly available benchmark datasets and outperformed the current state-of-the-art methods. It is anticipated that the proposed model could be helpful for both academia and drug discovery. Finally, a user-friendly web-server has been established and made freely available at: https://home.jbnu.ac.kr/NSCL/iN6-Methyl.htm.