ResSUMO: A Deep Learning Architecture Based on Residual Structure for Prediction of Lysine SUMOylation Sites

Abstract

Lysine SUMOylation plays an essential role in various biological functions. Several approaches integrating various algorithms have been developed for predicting SUMOylation sites based on a limited dataset. Recently, the number of identified SUMOylation sites has significantly increased due to investigation at the proteomics scale. We collected modification data and found the reported approaches had poor performance using our collected data. Therefore, it is essential to explore the characteristics of this modification and construct prediction models with improved performance based on an enlarged dataset. In this study, we constructed and compared 16 classifiers by integrating four different algorithms and four encoding features selected from 11 sequence-based or physicochemical features. We found that the convolution neural network (CNN) model integrated with residue structure, dubbed ResSUMO, performed favorably when compared with the traditional machine learning and CNN models in both cross-validation and independent tests. The area under the receiver operating characteristic (ROC) curve for ResSUMO was around 0.80, superior to that of the reported predictors. We also found that increasing the depth of neural networks in the CNN models did not improve prediction performance due to the degradation problem, but the residual structure could be included to optimize the neural networks and improve performance. This indicates that residual neural networks have the potential to be broadly applied in the prediction of other types of modification sites with great effectiveness and robustness. Furthermore, the online ResSUMO service is freely accessible.

An application of topological data analysis in predicting sumoylation sites

“…The characteristics of AAs in the protein fragment were shown to be an important parameter with which to predict sumoylation sites ( Khan et al, 2021 ) according to the large proportion of TF3. All of the features in TF4 were part of the optimal feature set, indicating that the features of the local region of peptide play a critical role in the recognition of sumoylation sites ( Zhu et al, 2022 ).…”

“…In general, computational approaches can be clustered into sequence-based and structure-based groups. Structure-based methods incorporate learning algorithms with sequence features derived from BLOSUM62 ( Xu et al, 2016 ; Zhu et al, 2022 ), pseudo amino acid composition (PseAAC) ( Jia et al, 2016 ), position relative incidence matrix (PRIM) ( Khan et al, 2021 ), etc. Structure-based methods integrate learning algorithms with physical features retrieved from local geometric information, such as half-sphere exposure (HSE) ( Sharma et al, 2019 ), backbone torsion angles, accessible surface area (ASA) ( Lopez et al, 2020 ), and contact number (CN) ( Dehzangi et al, 2018 ).…”

CNNArginineMe: A CNN structure for training models for predicting arginine methylation sites based on the One-Hot encoding of peptide sequence

“…Many classifiers for predicting various types of PTM sites have been developed by integrating machine-learning or deep-learning algorithms with different encoding features ( Chen et al, 2018a ; Huang et al, 2018 ; Chen et al, 2019 ; Lyu et al, 2020 ; Zhang et al, 2020 ; Zhao et al, 2020 ; Sha et al, 2021 ; Wei et al, 2021 ; Zhu et al, 2022 ). It has been found that the models based on deep-learning algorithms have better prediction performances than those based on traditional machine-learning algorithms.…”

Enhancing Sumoylation Site Prediction: A Deep Neural Network with Discriminative Features

“…Additionally, we divided the data randomly into 10 groups: 9/10 (i.e., 702 positives and 702 negatives) as the cross-validation dataset and 1/10 (i.e., 78 positives and 78 negatives) as the independent test dataset. The data frame sample function in the pandas Python library was used for random number generation [33].…”