The Study Of Tree Protein Secondary Structure Prediction Based On BP Neural Network

The study of tree protein secondary structure prediction is an important topic in the forest field.It has important academic significance,about further understanding of forest dynamics mechanism,improving forest resource information management and design of forest drug protection, and for researching on forest biology.The problems of tree protein secondary structure predicition were studied,in this paper.First,it was reviewed about research on tree protein structure predicition at home and abroad;then using four pure mathematical way to encoding forest protein to comparise;finally,get the forecast model using the good mathematical encoding of the mathematical way.The main research comtents and results are as follows:(1) Research progress on tree protein structure predicition at home and abroadSo far, it did not found using neural network to do the research on tree protein structure predicition.Tree protein structure was abtained by experimental methods(X-ray diffraction method and nuclearmagnetic resonance spetroscopy-NMR). Research focused on the genus Populus and rubber trees, as for the sandalwood genus, Pinaceae, Rosaceae, Ginkgo biloba, Eucommia slightly research.The first tree protein structure data was obtained on 1983-11-02 about electron transfer protein of black poplar.It was very slow about abtaining of tree protein structure,only an average of three each year.Overall, the research on tree protein structure is less ,and that has large development space.(2) Comparison research on 4 mathematical coding of amino acidAmino acid of tree protein was encoded using four common mathematical encoding methods-[-1 1]coding,five coding,orthogonal coding,21 coding,and then compared prediction accuracy by using BP neural network. The results showed that [-11] code is simple, understandable and the prediction accuracy is high compareing to other three kinds of coding,followed by 21 coding, orthogonal coding, five coding.(3)Prediction model of tree protein secondary structure based on BP neural network2600 amino acids extracting from 24 tree protein were carryed on prediction research in this article. Amino acids extracted were encoded by [-1 1] coding, then doing BP taining, through simulation results analysis, fitting and predictiong accuracy analysis,at last ,predicion models of tree protein secondary structure was obtained. The model overall prediction accuracy is 62.056%, for H it can achieve a high prediction accuracy of 72%, compared the same as previous mathematical encoding of protein secondary structure prediction ,the accuracy is higher.