Comprehensive assessment of three molecular serotyping methods of Salmonella based on genomic sequence

  Objective  To determine an alternative method for serotyping of Salmonella to replace the traditional serum agglutination test, we compared three molecular serotyping methods and selected a method with higher accuracy.

  Methods  A total of 509 isolates of Salmonella of 50 serotypes were used to extract DNA for whole-genome sequencing and predict the serotypes of the strains by online prediction using three methods: MLST, SalmonSeroPredicition and SISTR. Compared with the serotype identified by traditional serum agglutination, the consistency of the serotype prediction of three methods was analyzed to evaluate the accuracy of serotype prediction by each method.

  Results  The prediction accuracy rates of SISTR, MLST and SalmonSeroPredicition were 96.67%, 93.52%, 69.16%, respectively. The highest prediction accuracy rate of common Salmonella serotypes such as S. Indiana and S. Typhimurium was 100%, the accuracy rate was 99.17% for S. Derby and 95.74% for S. Enteritidis. The serotypes that all three methods predicted incorrectly were S. Enteritidis, S. Derby, Salmonella enterica subsp. salamae, subsp. arizonae and subsp. diarizonae etc. The main reasons for the error were the missing of sequence genes and the loss of the switching mechanism or the ability to express the flagellar gene.

  Conclusion  SISTR based on genomic sequence has high serotype prediction accuracy. In the case that traditional serum agglutination is unavailable or cannot be conducted in time, SISTR can replace serum agglutination test to determine the serotype of Salmonella.