Search

As one of post-translational modifications (PTMs), protein citrullination is crucial in a diverse array of cellular processes and implicated in a slew of human pathology. Therefore, accurate identification of protein citrullination sites (PCSs) is urgently needed to illuminate the reaction details and the complex pathogenesis related to the protein citrullination. In view of the limitations of the existing PCS predictors, this study proposes a novel and powerful sequence-based combined method named PCSPred_SC to further enhance the prediction performance. Various feature extraction methods are developed to mine sequence-derived biological information. Under the feature space, the predictive capabilities of different prediction algorithms, over-sampling methods, and feature selection methods are respectively explored. Experimental results indicate that the over-sampling methods are effective to solve the imbalanced dataset problem and the feature selection methods are significant in removing irrelevant and redundant features. On the same dataset using 10-fold cross validation, PCSPred_SC constructed by the combination of support vector machine (SVM), Adasyn, and t-distributed stochastic neighbor embedding (t-SNE) achieves much more outstanding performance than the competing methods, while reducing the number of features used for this task remarkably. It is anticipated that the proposed method will provide significant information to broaden our knowledge of citrullination-related biological processes.

As an unavoidable non-enzymatic reaction between proteins and reducing sugars, glycation can decline antioxidant defense mechanisms, damage cellular organelles, and form advanced glycation end products (AGEs), thereby resulting in a series of destructive physiological diseases. Identification and analysis of protein glycation sites will be beneficial to understand the complex pathogenesis related to the glycation. In this paper, a new glycation site predictor, DeepGly, is proposed based on a deep learning framework with a recurrent neural network (RNN) and a convolutional neural network (CNN). Firstly, for the class imbalance problem in the benchmark dataset, Long Short-Term Memory (LSTM) RNNs are designed to generate artificial peptides with glycation sites to form a balanced dataset. Then, the peptides in the balanced dataset are cleaved into a series of biological words, and continuous distribution representation is employed to transform the biological words into digital vectors. Finally, the digital vectors are input into the CNN with participations of the plurality and multiple convolution kernels to automatically extract various features, pooling layers to perform feature selection, and a softmax function to classify peptides. On the same datasets using 10-fold cross validation test, the prediction performance of DeepGly is far superior to that of existing methods, which indicates that the proposed method can be used as an ideal choice for protein glycation site prediction and also has a certain promotion effect on other related fields.

DNA replication is key to the inheritance of genetic information. Accurate, efficient and rapid identification of the origins of replication (ORIs) is crucial for understanding the mechanism of DNA replication. Especially for eukaryotes, each of their gene sequences contains multiple ORIs for more efficient replication. Although there are many predictors designed to identify eukaryotes’ ORIs, many of them are only targeted to the gene sequences with a fixed length. In addition, the prediction accuracies are not satisfying, which still has great room to be improved. In view of the limitations in this field, a convolutional neural network-based approach is developed in this study to identify ORIs with different lengths in Saccharomyces cerevisiae (S. cerevisiae). As combining this study with the field of Natural Language Processing (NLP), trinucleotide feature vectors are constructed by Word2vec to represent each trinucleotide so as to the subsequent ORIs identification using Text-Convolutional Neural Network. As a result, the overall success rate of 88.3% was achieved which proved the effeciency of the proposed method to identify ORIs with any length.

Accumulation of the final product during glycation reaction often leads to many diseases, such as diabetes, Alzheimer’s disease and atherosclerosis. Identifying the glycation site can help researchers to understand the pathogenesis and provide new ideas on how to treat these diseases. In this paper, we develop a new predictor by using the support vector machine which apply four feature extractions to encode peptide chains such as binary code sequence, grey incidence degree, accessible surface area and secondary structure probability. The maximum relevancy minimum redundancy (mRMR) feature selection algorithm is used to select the optimal 170 features for the prediction problem. In training set, the performance of Gly-Predict is assessed with an accuracy of 84.815%, a sensitivity of 80.156%, a specificity of 88.868%, and a Matthews’s correlation coefficient (MCC) of 68.411% by k-fold cross validation (k = 5). To objectively evaluate Gly-predict, we tested our model on an independent dataset and compared with previous predictor. The results indicate that Gly-predict is superior to existing glycation site predictors.