Research On Linear B-cell Epitope Prediction Model

B-cell epitope is an fragments of antigen molecule combined with antibody. The continuous amino acid sequence of the antigen epitope called linear B-cell epitopes. Linear B-cell epitope prediction research has important application value. Detection have two ways:experimental methods and computational prediction methods. Experimental methods as time-consuming, the consumption of human and other shortcomings can’t meet the actual demand. Because of the repeatable and timesaving computational prediction method has become an inevitable choice.Using the calculation method of linear B-cell epitope prediction study is divided into two phases, early studies used only to predict the physical and chemical properties of amino acids, the prediction accuracy is generally low. In recent years, with the linear B-cell epitope of data increases, the emergence of a variety of epitope libraries, using machine learning methods to predict the way to become a research hotspot. This paper is about the linear B-cell epitope prediction model research, the main work is as follows:1. Common research on linear B-cell epitope prediction model. Deeply analysis the characteristic of the linear B-cell epitopes,then analysis of the current commonly used feature extraction methods and prediction methods based on machine learning based on reviewing the relevant literature study and finally given the relatively accepted prediction model validation methods and evaluation criteria. Prepare for the further research work.2. Propose a weighted bayes linear B-cell epitope prediction model. The model based on bayes feature extraction method and introduce the concept of amino acid pair antigenicity scale, effectively extract information of composition and structure of the amino acid sequence and enhance sequences of class correlation. In BCPred, Chen datasets tests show that the improved model can improve certain prediction. Compared with other methods, this method can effectively improve the prediction accuracy.3. Proposed based on multi-feature fusion linear B-cell epitope prediction model. In order to further improve the prediction accuracy, using serial fusion method to increase the information of the feature characteristics. The main features include a combination of amino acid sequence order, consisting of information and physical and chemical properties. According to the different effect on the classification between the various dimensions of feature of the combination improved linear discriminant analysis and dimensionality reduction. Combine SVM technology to get the final linear B-cell epitope prediction model. Our model combines various characteristics of the linear B-cell epitope, increases the amount of the feature information, and use linear discriminant analysis reducing the redundancy of the feature information. In the five test data sets show that compared to other methods, this prediction model has been further improved.