Protein nitration and S-nitrosylation are important post-translational modifications induced by reactive nitrogen species (RNS) compounds.
By utilizing deep learning, DeepNitro has developed a new algorithm for the prediction of protein modification sites.
DeepNitro can simultaneously predict the exact locations of Tyr-nitration, Trp-nitration, and S-nitrosylation.
Advanced features are encoded by means of modified PSSM and k-space scheme.
Deep neural network is utilized to extract high dimension information.
DeepNitro outperforms other tools in predicting Tyr-nitration, Trp-nitration, and S-nitrosylation sites.
Model of DeepNitro is constructed by a deep neural network with seven fully-connected layers.
To avoid overfitting, the dropout and regularization are applied in the model.
Softmax function is implemented in the classification layer.
To assist further functional analysis, DeepNitro implements an automatic pipeline for visualizing the prediction results.
DeepNitro can present the graphical representation of the inputted proteins together with their predicted sites and domain organization in the visualization panel.