TransCNN: A novel architecture combining transformer and TextCNN for detecting N4-acetylcytidine sites in human mRNA

Abstract

N4-acetylcytidine (ac4C), a pivotal post-transcriptional RNA modification, is central to understanding transcriptional regulation and diverse biological processes. As a key determinant of RNA structural stability and functional regulation, ac4C has been strongly associated with multiple human diseases. We can obtain a better understanding of regulation mechanism of gene expression by identifying ac4C sites rapidly and precisely. However, existing predictive approaches are constrained by limitations in feature representation and sequence context modeling, necessitating the development of advanced methodologies. In this study, we introduce a novel architecture named TransCNN that integrates transformer and Text convolutional neural network (TextCNN) to predict ac4C sites. TransCNN demonstrates superior performance compared to existing models on both 10-fold cross-validation and independent dataset with the accuracy of 83.27 % and 82.89 %, respectively. The enhanced performance of TransCNN is attributed to the transformer's ability to extract adaptive features and TextCNN's capability to form both narrow and broad connections within the sequence. This study aims to contribute significantly to the field by advancing the understanding and prediction of RNA modifications. The datasets and code used in this study are available at https://github.com/liukai23157/TransCNN.