DeepGlut: A Deep Learning Framework for Prediction of Glutarylation Sites in Proteins

Glutarylation Post Translational Modifications (PTM) means the enzymatic or chemical modifications in lysine residue of a protein molecule after the translation step. These modifications play a vital role in various kinds of biological and physiological processes which might be occurred by the addition of a glutaryl group to a lysine residue. It also causes several diseases. It is a type of protein modification in both eukaryotes and prokaryotes cells. Here, a question arises: which one can be glutarylated, and which one cannot be glutarylated of these protein residues. The answer will help us to understand the mechanism of glutarylation that can also be beneficial for the development of new drugs. In this study, a deep learning tool named DeepGlut is developed to identify glutarylation sites based on a Convolutional Neural Network(CNN). For feature extraction, one hot encoding is incorporated to encode glutarylation sites. To minimize the imbalanced problem in the glutarylation training dataset, the over-sampling technique has been used which is further evaluated by 10-fold cross-validation, DeepGlut achieves a satisfying performance with a sensitivity of 0.8419, a precision of 0.7267, an accuracy of 0.7616 and an Area Under Curve or AUC of 0.8383. Our proposed predictor DeepGlut can be downloaded from https://github.com/urmisen/DeepGlut.