Multi-layer Ensemble Classifiers on Protein Secondary Structure Prediction
ICIC '08 Proceedings of the 4th international conference on Intelligent Computing: Advanced Intelligent Computing Theories and Applications - with Aspects of Theoretical and Methodological Issues
IEEE Transactions on Neural Networks
Learning without local minima in radial basis function networks
IEEE Transactions on Neural Networks
Hi-index | 0.00 |
Protein structure prediction plays an important role in the expression and analysis of the protein sequences. As the high performance analytical instruments will need a high cost, so the way through the algorithm to get the secondary structure information has become a very important method. The traditional secondary prediction is directly based on comparing the sequences, which relied solely on the Hamming distance to classify them. Analyzing this data organization method, it not only inconsistent with the practical problems, but also throws away a lot of information. In the end the accuracy is often too low. Research shows that each amino acid can be seen as a force with its own size and direction. With the effect of all amino acids force, they form the spatial structure of the protein sequence. And then we can describe the distortions in the way similar to the description of the force diagram. It is difficult to determine whether the amino acid sequence in a forward or a reverse order, so the symmetrical position of amino acids will have a similar effect to the middle area. Based on the above analysis, we use a symmetrical binding form of organization to construct a new input sample. When testing the accuracy we use the neural network method. Because the force in different positions is different, so we add a property factor to the neural network model. Then the power of amino acids in different positions can be adjusted by the property factor. In the end we can not only get the most suitable parameters to get a high accuracy, but also reveal the relations between the different amino acids.