Research Of Prediction Method For Lysine Post-translational Modification Sites

In the post-genome era,proteomics is such an important research content.Protein is directly involved in the lives of human activity.A common protein will become a functional mature protein through post translational modifications(PTMs).However,the use of experimental determination to recognize PTMs sites is usually limited.Machine learning methods has gradually become a research hotspot.In this paper,two corresponding prediction methods of glycation and SUMOylation were constructed for the modification of lysine sites.The detailed research content is as follows:(1)Protein gly cation is related to many human diseases such as age-related diseases and the development of diabetes.PredGly,a novel predictor,is presented to help researchers filter supposed sites prior to experimental verification so as to improve the efficiency of experimental detection.PredGly extracts three types of feature expressions,including sequence information,physical and chemical property,and evolutionary information.In addition,XGboost is used to delete irrelevant and redundant information from fusion features.Based on a new set of independent tests,PredGly also performs the other lysine glycation prediction tools.(2)SUMOylation not only plays an important role in regulating a variety of cellular processes,but is also closely associated with serious human diseases.DeepSUMO,a deep convolutional neural network(CNN),is presented for the identification of general lysine SUMOylation sites.DeepSUMO applies convolution neural network(CNN)to capture sequence information from the original protein sequences combining the enhanced amino acid composition.In this case,predictor is a multi-input model based on two kinds of sequence information.In order to compare with different traditional machine learning algorithms,we use the other four classifiers with the cross validation.