Incorporating convolutional neural networks and sequence graph transform for identifying multilabel protein Lysine PTM sites

Protein post-translational modification (PTM) is a process where proteins, after being created, are modified through chemical processes in the body. Some recent studies have shown that PTM sites play an important role in signaling transduction, transcriptional regulation, and apoptosis. Among different types of PTM, the modification at Lysine (K) is the most frequently observed PTMs. Therefore, identifying Lysine PTM sites could be the key to decipher its mysterious structures and functions which are important in cell biology and diseases. Few studies have addressed this necessary problem using computational models; however, the predictive performance is not satisfactory. Thus, we aim to improve the performance results by using a novel combination with convolutional neural networks and sequence graph transform. The absolute-true rates within the cross-validation and independent achieved 85.21% and 85%, respectively. Compared to other methods as well as state-of-the-art published works, our proposed model reach performed better on a benchmark dataset. Our results show that we can propose an efficient model for improving the predictive performance of Lysine PTM sites. Moreover, it also suggests that deep learning and graph theory-based features could open a new avenue in biochemical modelling using sequence information.