Study On Protein Secondary Structure Prediction Based On Neural Network

The study of protein structure and function is an important topic in the field of bioinformatics.The prediction of protein secondary structure is one of the key tasks in the study of protein structure and function.In recent years,deep learning methods have been widely used in protein secondary structure prediction field,but there are problems such as long training time and low parallelism.The combined application of ensemble learning and deep learning techniques to protein secondary structure prediction can reduce the cost of model training.At the same time,different models can be combined to analyze amino acid information from multiple perspectives,so as to improve the prediction accuracy of the model.Therefore,a protein secondary structure prediction model was constructed based on deep learning method combined with ensemble learning technology in this paper.The main works of this paper are as follows:1.A hybrid model of CNN and Bagging technology is constructed.The model uses CNN instead of the simple SVM weak classifier to effectively improve the prediction accuracy of protein secondary structure.At the same time,the combination strategy of Bagging model was optimized,and the prediction accuracy of protein secondary structure was finally improved from 68.1% to 71.9%.2.A hybrid model of neural network and Stacking technology is built.On the basis of CNN and Bagging mixed model,combined with Stacking technology,the prediction accuracy of the model is further improved,reaching 73.4%.At the same time,it is further proposed that the data set can be divided according to the protein length and the results of 3 classification and 8 classification can be combined to improve the prediction accuracy of the model,and the prediction accuracy of the model finally reaches 75.6%.3.The translation model of protein secondary structure was constructed.The prediction of protein secondary structure was turned into a translation problem.N-gram algorithm was used for word segmentation,and an end-to-end translation model was constructed to predict protein secondary structure.Finally,the prediction accuracy of protein secondary structure was the highest of 70.6%.