Toward prediction of multi-states secondary structures of protein by neural network

抄録

Usually, the prediction of protein secondary structure by a neural network is based on three states (α-helix, β-sheet and coil). However, a recent report of protein of which structure is determined presents more detailed secondary structure as 3₁₀-helix. It is expected that more detailed secondary structure of protein should be predicted. In application of neural network to the prediction of multi-states secondary structures, some problematic points are discussed. The prediction of globular protein secondary structures is studied by a neural network. The application of a neural network with a modular architecture to prediction of protein secondary structures (α-helix, β-sheet and coil) is presented. Each module is a three layer neural network. The results from the neural network with a modular architecture and with a simple three layer structure are compared. Overlearning effect is investigated in ordinary and modular neural networks. The prediction accuracy by a neural network with a modular architecture is higher than of the ordinary neural network. The 3, 4 and 8 state classification scheme of secondary structures are considered in the ordinary three layer neural network. The percentage of correct prediction depends on these state classification method. Furthermore, for 3 and 4 state classification scheme of protein secondary structures, the consistencey of outputs of modules on the neural network with modular architecture is investigated.