Research On Protein Classification Prediction Based On Deep Learning

Proteins are the products of gene expression and the basic elements in life.They play an important role in maintaining the life activities of organisms,which makes proteomics become an important research field in the post-genomic life sciences.It is of great significance to accurately predict and classify proteins for the study of their structure and function.With the development of the post-genome era,protein data has proliferated,and it is time-consuming and costly to determine the protein category through traditional biological experiments.Therefore,it is necessary to develop a method for protein prediction using theoretical methods and computational techniques.Based on deep learning algorithm,this paper selects two types of disease-related proteins: clathrin and DNA-binding protein for research,then establishes two protein prediction models respectively,and makes online discriminatory analysis for them.The specific content of this paper is as follows:1.Prediction of clathrin based on hybrid deep learning model.Clathrin is a connectin,and clathrin-mediated endocytosis is extremely important in life activities.In addition,clathrin deficiency also affects the occurrence of many diseases.Therefore,the identification of clathrin is of great significance for maintaining the health of living organisms.Here,we used the original sequence information of clathin,employed the grouping of physical and chemical properties of amino acids to encode the features,and introduced the convolutional neural network and long-short time memory neural network to construct a hybrid prediction model Deep CLA.The results of cross-validation and independent test show that the hybrid depth model can effectively improve the prediction performance,and the prediction performance of Deep CLA is better than that of existing forecasting tools.This also provides clues for further analysis the structure and function of clathrin.2.DNA-binding protein prediction based on deep residual network.DNA-binding proteins bind to DNA and affect DNA transcription,replication and selective splicing in humans.We collected all the DNA-binding protein data and encoded the protein sequences based on the composition of amino acid pairs in k-space,so that the model could extract effective features.A new model DeepDBPNet was constructed by introducing residual blocks in the Deep residual network.By analyzing the experimental results and comparing them with previous predictive tools,the model further improved the classification performance of DNA-binding proteins.