Objective: To analyze the bioinformatics of domestic Clostridium difficile toxin B(TcdB) and prepare it to provide data support for the development of effective vaccines. Methods: Using bioinformatics software such as Snippy, Blast, Muscle, and the dist.alignment() and hclust() functions in R, 1 355 strains of Clostridium difficile from NCBI GenBank in China were compared and analyzed, and TcdB were grouped. The maximum likelihood tree and phylogenetic tree were beautified and displayed using iTOL. An online bioinformatics analysis website was used to predict and analyze the spatial structure and antigenic epitopes of the two largest subgroups, TcdB1 and TcdB2. The antigen protein TcdB was expressed and purified by prokaryotic system. Results: According to the genotype of toxin B, the 1 355 prevalent strains of Clostridium difficile in China could be roughly divided into 12 subtypes, among which TcdB1 and TcdB2 were the main subtypes, accounting for more than 93.94% of all isolated strains, and about 17.20% of the strains were nontoxigenic or lack TcdB. The antigen epitope prediction of TcdB1 and TcdB2 showed that their antigen epitope distributions were basically the same, and many of them were distributed outside the C-terminal combined repetitive oligopeptides domains. Conclusion A specialized typing system for C. difficile TcdB in China has been established, and its main subtypes have been predicted for antigenic epitopes. The screened TcdB has been expressed for recombinant preparation.