| Summary: | An important role is played by N6-methyladenosine (m6A) in RNA methylation modification. The modification information is crucially required for development in the field of medicine. Biochemical experiments for m6A identification have demonstrated high-quality results. However, this process is not a feasible solution due to its cost and time constraints. Recently, artificial intelligence has played a role to produce effective models which can be utilized for speedy and efficient identification of N6-methyladenosine sites. In consensus, a computational model named as m6A-NeuralTool is proposed in this study. The m6A-NeuralTool makes the final prediction for the identification of m6A sites by applying majority voting on three different sub-architectures. These sub-architectures uses a set of Convolution layers to extract the important features from the one-hot encoded input sequence. Further, one of the sub-architectures uses fully connected layers for the classification while the other two uses Support Vector Machine and Naive Bayes. The proposed m6A-NeuralTool is evaluated on four different species datasets using 10-fold cross-validation, where the proposed tool outperformed the existing state-of-art models for modification identification of m6A sites. When compared with existing models the proposed model increased the accuracy by 14.8%, 8.8%, 6.3%, and 4.9% for the datasets of Saccharomyces cerevisiae (S. cerevisiae), Arabidopsis thaliana (A. thaliana), Mus musculus (M. musculus) and Homo sapiens (H. sapiens) species respectively. The proposed model will allow the researchers from the field of bioinformatics and medicine to accurately identify the modification in m6A sites and use the information in the development of different products which would be beneficial for the community. The m6A-NeuralTool can be accessed freely at: http://nsclbio.jbnu.ac.kr/tools/m6A-NeuralTool/.
|
|---|
