Dissertation > Excellent graduate degree dissertation topics show

Hierarchical Neural Networks for Protein Secondary Structure Prediction

Author: LiWei
Tutor: ChenYueHui
School: Jinan University
Course: Applied Computer Technology
Keywords: Psi-Blast compare protein structure prediction Position Special Scoring Matrix (PSSM) Hierarchical Radial Basis Function (HRBF) neural network
CLC: TP183
Type: Master's thesis
Year: 2009
Downloads: 79
Quote: 0
Read: Download Dissertation

Abstract


The progress by Amino acid sequence to predict protein secondary structure is called protein secondary prediction. The Amino Acid (AA) sequence has different length, and they often with different AA order. The difference of them will decide the different structure. To study the structure is very important, which not only help us to understand the use of protein, but also help us to know how the protein to exert the function of biology. It’s very important to the biology, physic and medicine to know the reciprocity of proteins.The gap of the number between the known structure and the unknown structure is becoming larger and larger, after the finishing of the genome plan in 2003. So the structure prediction is becoming more and more pressure.The aim of this article is how to construct a model to predict protein secondary structure, which can improve the correct ratio effectually. The article is composed by three different parts. First of all we will distill the important character from the AA sequence. Than we choose a fitful arithmetic to optimize our neural network. Finally the structure of the neural network will be constructed.1. Distill the character from the AA sequence. In order to predict the structure of the AA sequence, first of all we must distill the useful character of AA sequence. It’s very important about which character will be used for the prediction. Different distill method will get different character. Now we can distill character from signal AA, AA sequence and the sequence with the same configuration. In this article we use Psi-pred method to distill the same configuration sequence character. By experiment we can know, this method can get higher predict rate, and also has deeply mathematics theory.2. Structure prediction. In fact, the structure is calculated by the character. We use the character to analysis the rule of the hidden, and then we can get the structure of sequence by the rule. The AA sequence is a very big database, so calculating the rule by computer often needs a lot of time. But the neural network can help us to solve this question, which can learn the rule effectively in very short time and predict the protein structure.The very important aspect of the neural network is the choosing of optimal arithmetic, which can decide the efficiency the neural network. Often using different arithmetic can get different efficiency and different prediction rate. In this article we will compare different optimal arithmetic, and then choose the best arithmetic to use in our prediction structure. In the process of the prediction, we found the number of the AA with different structure is very different, which can cause the imbalance training problem. The bagging method in computational aptitude can solve the imbalance problem, so we use the method to reform the initial AA database. The results tell us the method is very efficiency in this question. In traditional method a two layered neural network is often used, in this article we use more layered to predict and transform the multi problem to binary problem, then ensemble the binary problem to multi problem. A lot of experiments of them tell us these methods are very efficiency. In traditional method, the neural networks structure is designed by experience, which is difficult to be accepted by us as a deep theory. So in this article we use Hierarchical Radial Basis Function (HRBF) neural network to design the prediction structure. HRBF neural network not only can optimal the weight of the network, but also can help us design the structure of the neural network, and another very important character is the choosing of input character. The using of HRBF help us improve the prediction rate and reduce the time of computation.

Related Dissertations

  1. Based on the feature fusion membrane protein transmembrane helix prediction,Q51
  2. Paralleling Genetic Annealing Algorithm with OpenMP,TP301.6
  3. Predictions of Human Kinome-Small Molecule Interactions,Q55
  4. Full atomic encoding proteins predicted solvent accessibility,Q51
  5. The Study on Atomic Distance-Dependent Pair-wise Statistical Potentials in Protein Structure Prediction,Q51
  6. Study of Particle Swarm Optimization for Protein Structure Prediction,TP301.6
  7. Cloning and Expression of Estrogen Related Receptor Gene of the Polyrhachis Vicina Roger and the Apis Mellifera,Q78
  8. Breeding of ε-Poly-L-Lysine-Producing Strain and Studies on Fermentaion Process,TQ922
  9. Heuristic Algorithm for Circle Packing Problem and Structure Prediction Problem of Model Protein,TP301.6
  10. Application of Simulated Annealing Algorithm in Prediction of Protein Structure,TP301.6
  11. Research on Prediction of Protein Contact Maps,Q811
  12. The reality of protein structure prediction method for solving,Q51
  13. Heuristic Algorithms for the Protein Structure Prediction Problem and the Rectangles Packing Problem,Q51
  14. Research on Several Problems of Protein Structure Prediction and Its Software Realization,Q51
  15. Study on Models of Protein Structure Prediction,Q51
  16. Quasiphysical and Anthropopathic Algorithm for Protein Structure Prediction,TP301.6
  17. A Triangulated HP Model of Protein Folding Structure Prediction and Algorithm,TP301.6
  18. Protein Structure and Function Research Method of,Q51
  19. Quasiphysical and Anthropopathic Algorithm for Protein Structure Prediction,TP301.6
  20. Isolation and Expression Analysis of the NTLEAFY and NTA Gene in Narcissus Tazetta Var. Chinensis Roem,S682.21

CLC: > Industrial Technology > Automation technology,computer technology > Automated basic theory > Artificial intelligence theory > Artificial Neural Networks and Computing
© 2012 www.DissertationTopic.Net  Mobile