iSS-CNN: Identifying splicing sites using convolution neural network

RNA splicing is an important post-transcriptional modification of eukaryotic organisms in which a single gene can code for different proteins that have different biological functions. Thus, accurate identification of RNA splicing sites sequences is important for both drugs discovery and biomedical research. However, through laboratory techniques the discrimination of the splicing sites is very expensive. Therefore, an accurate computational model is needed. In this work, we introduce an efficient convolution neural network (CNN) model called iSS-CNN for splicing sites identification. Previous methods utilized hand-crafted features for identifying splicing sites, however, the proposed model extracts the features of the splicing sites automatically using the proposed CNN model. The performance of iSS-CNN has been evaluated on benchmark datasets and produced better outcomes than the existing methods. The iSS-CNN predictor obtained 96.66% of accuracy for a dataset containing splicing donor sites (SDS) and 93.57% of accuracy for a dataset containing splicing acceptor sites (SAS) using 5-fold cross-validation test. A webserver for the iSS-CNN tool has been established and made available at https://home.jbnu.ac.kr/NSCL/iss-cnn.htm.