iSUMOK-PseAAC: prediction of lysine sumoylation sites using statistical moments and Chou’s PseAAC

Abstract

Sumoylation is the post-translational modification that is involved in the adaption of the cells and the functional properties of a large number of proteins. Sumoylation has key importance in subcellular concentration, transcriptional synchronization, chromatin remodeling, response to stress, and regulation of mitosis. Sumoylation is associated with developmental defects in many human diseases such as cancer, Huntington’s, Alzheimer’s, Parkinson’s, Spin cerebellar ataxia 1, and amyotrophic lateral sclerosis. The covalent bonding of Sumoylation is essential to inheriting part of the operative characteristics of some other proteins. For that reason, the prediction of the Sumoylation site has significance in the scientific community. A novel and efficient technique is proposed to predict the Sumoylation sites in proteins by incorporating Chou’s Pseudo Amino Acid Composition (PseAAC) with statistical moments-based features. The outcomes from the proposed system using 10 fold cross-validation testing are 94.51%, 94.24%, 94.79% and 0.8903% accuracy, sensitivity, specificity and MCC, respectively. The performance of the proposed system is so far the best in comparison to the other state-of-the-art methods. The codes for the current study are available on the GitHub repository using the link: https://github.com/csbioinfopk/iSumoK-PseAAC.

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

“…We collected 47,752 experimentally verified SUMOylation sites on 8687 human proteins from the literature and the PLMD database [ 11 , 20 , 28 , 29 ] (see Table S2 for details). After CD-HIT clustering, 37,273 SUMOylation sites of the 5851 protein representatives as positive samples (See Methods for details).…”

“…Specifically, we collected 47,752 experimentally verified SUMOylation sites of 8687 proteins from the human proteome. These data are derived from either the literature or databases ( Table S2 ) [ 11 , 20 , 28 , 29 ]. These proteins were grouped using the CD-HIT tool [ 28 , 29 ] with a sequence identity of 40%, and 5851 clusters were generated.…”

“…By contrast, computational approaches for predicting SUMOylation sites have attracted considerable attention because of their convenience and efficiency. They include pSumo-CD [ 13 ], iAcet-Sumo [ 14 ], SUMOgo [ 15 ], SumSec [ 16 ], HseSUMO [ 17 ], SUMO-Forest [ 18 ], C-iSUMO [ 19 ], and iSUMOK-PseAAC [ 20 ], which are summarized in Table S1 . They can be clustered into two groups: structure-based, which includes three classifiers, and sequence-based, which covers the remainder.…”

“…SUMO-Forest is a method based on cascade forest [ 24 ] with the features of bi-gram, k-skip-bigram, and statistics property [ 25 ]. The deep learning subgroup includes iSUMOK-PseAAC, an ANN (artificial neural network)-based classifier with the features of statistical moments and PseAAC [ 20 ]. The iSUMOK-PseAAC model is the latest in these reported classifiers and compares favorably to other classifiers [ 20 ].…”

“…The deep learning subgroup includes iSUMOK-PseAAC, an ANN (artificial neural network)-based classifier with the features of statistical moments and PseAAC [ 20 ]. The iSUMOK-PseAAC model is the latest in these reported classifiers and compares favorably to other classifiers [ 20 ]. The deep-learning algorithm, different from the traditional machine-learning algorithm that requires pre-defined informative features, has the strong capability to learn sparse representation in a self-taught manner and can auto-capture the most informative features.…”