EMDL-ac4C: identifying N4-acetylcytidine based on ensemble two-branch residual connection DenseNet and attention

Introduction: N4-acetylcytidine (ac4C) is a critical acetylation modification that has an essential function in protein translation and associated with number of human diseases. Methods: The process identifying ac4C sites by biological experiments too cumbersome costly. And the performance several existing computational models needs to be improved. Therefore, we propose new deep learning tool EMDL-ac4C predict sites, which uses simple one-hot encoding for unbalanced dataset using downsampled ensemble network extract important features identify sites. base learner this model consists modified DenseNet Squeeze-and-Excitation Networks. In addition, innovatively add convolutional residual structure parallel dense block achieve effect two-layer feature extraction. Results: average accuracy (Acc), mathews correlation coefficient (MCC), area under curve Area on ten independent testing sets are 80.84%, 61.77%, 87.94%, respectively. Discussion: Multiple experimental comparisons indicate outperforms predictors it greatly improved predictive At same time, could provide valuable reference next part study. source code data available at: https://github.com/13133989982/EMDLac4C .