iKcr-DRC: prediction of lysine crotonylation sites in proteins based on a novel attention module and DenseNet

Lysine crotonylation (Kcr) is a recently identified post-translational modification that predominantly occurs on lysine residues and plays a crucial role in regulating gene expression, cellular metabolism, and various biological processes. Increasing evidence has linked Kcr to the pathogenesis of major diseases such as cancer, highlighting the importance of accurately identifying Kcr sites for understanding disease mechanisms and normal cellular function.

In this study, we present a novel deep learning-based computational model, named iKcr-DRC, for the accurate prediction of lysine crotonylation sites. The model leverages a densely connected convolutional network (DenseNet) as its backbone to effectively capture high-level local features from protein sequences. Additionally, we introduce an enhanced channel attention mechanism with a short-circuit connection design, endowing the network with residual properties and improved feature refinement capabilities.

The experimental results show that the iKcr-DRC model achieves 90.30%, 78.35%, 84.33% and 69.15% for sensitivity, specificity, accuracy, and Matthew's correlation coefficients, respectively. These results indicate a significant improvement over existing state-of-the-art Kcr prediction tools.

The proposed iKcr-DRC model provides an effective and innovative approach for predicting lysine crotonylation sites. It holds great potential for advancing applications in bioinformatics and enhancing the understanding of protein post-translational modifications. An online prediction tool based on the iKcr-DRC model is freely accessible at: http://www.lzzzlab.top/ikcr/.