Insulinoma associated protein-2 (IA-2) is a major autoantigen in type 1 diabetes. Epitopes of IA-2 autoantigens are often intracellular and only a few domains can be recognized by B lymphocyte. Conserved epitopes on IA-2 have been determined previously but other epitopes are not well defined. In this study, we use three computational methods, PepSurf, Pep-3D-Search and SEPPA to predict IA-2 epitopes. We concerned IA-2 PTP protein domains and input the crystal structure of IA-2691–977 (PDB ID: 2I1Y) to the three softwares.PepSurf predicted 2 clusters while the best cluster showed IA-2719-747were epitopes of IA-2. Pep-3D-Search predicted 3 clusters. The Cluster 3 contained 13residuesIA-2843-855.SEPPApredicted a total 45 residues includingresiduesIA-2887-897. We selected IA-2719-747，IA-2843-855and IA-2887-897 as candidate B lymphocyte epitopes of IA-2 after computational analysis. Mutants of the three candidateepitopes were constructed using site-directed mutagenesis.Binding with serum autoantibodies from IA-2 autoantibody positive type 1 diabetic patients was inhibited by mutations of these residues. In conclusion, softwares such asPepSurf, Pep-3D-Search andSEPPA can be used to predict autoantigen IA-2 epitopes of B lymphocyte. Moreover,IA-2719-747, IA-2843-855andIA-2887-897are new IA-2 epitopes of B lymphocyte.