Exploring the Transcriptional and Translational Features Using Deep Neural Networks for mRNAs Classification

Recent advent of the second and third generation of sequencing has uncovered many novel transcripts. These novel transcripts could have crucial functions in different biological processes and might be related to challenging diseases and pathogenesis. However, whether these genes should be classified as protein coding RNAs (pcRNAs) or long non-coding RNAs (lncRNAs) is still debated and unclear. In this study we propose a coding potential classification framework based on deep neural networks and novel features from RNA-seq and Ribo-seq data to classify RNAs transcripts into protein coding and long non coding. As far as we know, this is the first method that uses RNA-seq and Ribo-seq as predictors to classify RNAs using a deep neural network model. Compared to other methods, the prediction of our method reached 97.4% accuracy.