Intelligent Computation In The Research On The Application Of Protein Structure Prediction

During the post-genomic era, it is extremely significant to study the spatial structure ofthe protein.It would offer us several projects of the early diagnosis of disease and of theprecise drug screening after studying the structure of protein, then we can make goodunderstanding of the relationship between structure and function of it.Protein secondarystructure prediction is an important part of the entire3D structure prediction.It plays animportant role to study the secondary structure when we study the structure and function ofprotein.The key to the problem of the secondary structure prediction is to establish thecorresponding relationship between the protein secondary structure model and andamino acidsequence.This article predict the protein secondary structure through the use of neural network. Neural network is a widely used model for protein secondary structure prediction while it is defective in its algorithms. It is easy to fall into the problem of local minima and the slow convergence rate.Then this can lead to low accuracy rate and it is difficult to achieve a good performance.In order to improve the accuracy rate of protein secondary structure predictio,this article uses the method of optimization of the neural network which is including the genetic algorithms and particle swarm optimization to make the result perfect.We design a complicated neural network model due to the old flawed single neural network to improve the prediction of neuralnetwork for the protein secondary structure.In this article, the specific research include the following four points.Based on BP neural network,the article trys to predict the protein spatial structure and studys the different sequence extration method,but found that the prediction accuracy of the BPnetwork is not very satisfactory. Therefore,it achieves certain effects when we attempt to improve BP algorithm.At the same time,it takes advantage of the PSO algorithm to optimize BPalgorithm, the algorithm successfully resolved the problem of local minima and slow convergence speed of the model.then it effectively improve the accuracy of the prediction.In order tofurther improve the prediction accuracy, this paper applys genetic algorithm to improve thefunction of the neural network gradient descent method. A genetic algorithm to improve BPnetwork prediction model of the structure has been put forward,and it proceeds the simulated experiment on25PDB data sets.Ultimately, the model prediction accuracy reaches about75%.Owing to the low accuracy of the traditional single neural network prediction,5BPnetworks have been used to design a two-tier complicated network structure model, and thetraditional method of encoding has been improved either. The simulation results show thatthis complex network model has been greatly improved the prediction accuracy compared to asingle neural network, the final prediction accuracy is close to78%, which also verified thatthe complex neural network has more advantages over a single neural network in theprediction problem of secondary spatial structure of protein。This article shows that the intelligent calculation is very effective to proceed the protein secondary structure prediction,especially after the optimization of neural network algorithm and structure, the accuracy rate of the whole model is very perfect.