Background: An epitope is a specific portion of a macromolecular antigen that can determine
antigen specificity, and has great significance in studying adaptive immune responses. It can be a
linear fragment in the antigen structure (also called a linear B-cell epitope) or an area of conformational
structure in space (also known as a conformational B-cell epitope). However, the methods of
empirical testing used to identify epitopes are costly and time consuming.
Objective: The objective of this study is to provide an efficient predictor for distinguishing linear
Method: In this study, we present a predictor model based on the incorporation of information on the position-
specific amino acid propensity, composition of amino acids, composition of pairs of amino acids
and position-specific pair of amino acids propensity. And F-Score was used to select valid features.
Results: In jackknife cross-validation, our model achieved an overall sensitivity of 92.59%, specificity of
95.47%, accuracy of 94.36% and Matthews correlation coefficient of 0.8729 on a non-redundant dataset.
Conclusion: The results confirm the constructed model is superior to other existing methods.