A Prediction Method of Protein Contact Based on Information Granules

In this paper, a prediction strategy of protein contact on the basis of information granules and the method to realize it have been brought forward. By this method, this study chooses some properties concerning protein 3D structure to encode, which are used as training samples of prediction model. On the basis of information granules, after dividing the training samples into several groups by applying the way of clustering analysis, this study makes use of neural network to construct a prediction model of protein contact. The experiment result indicates that this method can not only increase the prediction accuracy of protein contact map, but also realize incremental study in terms of the new training sample.