Statistical superiority of genome-probing microarrays as genomic DNA-DNA hybridization in revealing the bacterial phylogenetic relationship compared to conventional methods.

The genomic DNA-DNA hybridization (DDH) method has been widely used as a practical method for the determination of phylogenetic relationships between closely related biological strains. Traditional DDH methods have serious limitations including low reproducibility, a high background and a time-consuming procedure. The DDH method using a genome-probing microarray (GPM) has been recently developed to complement conventional methods and could be used to overcome the limitations that are typically encountered. It is necessary to compare the GPM-based DDH method to the conventional methods before using the GPM for the estimation of genomic similarities since all of the previous scientific data have been entirely dependent on conventional DDH methods. In order to address this issue we compared the DDH values obtained using the GPM, microplate and nylon membrane methods to multi-locus sequence typing (MLST) data for 9 Salmonella genomes and an Escherichia coli type strain. The results showed that the genome similarity values and the degrees of standard deviation obtained using the GPM method were lower than those obtained with the microplate and nylon membrane methods. The dendrogram from the cluster analysis of GPM DDH values was consistent with the phylogenetic tree obtained from the multi-locus sequence typing (MLST) data but was not similar to those obtained using the microplate and nylon membrane methods. Although the signal intensity had to be maximal when the targets were hybridized to their own probe, the methods using membranes and microplates frequently produced higher signals in the heterologous hybridizations than those obtained in the homologous hybridizations. Only the GPM method produced the highest signal intensity in homologous hybridizations. These results show that the GPM method can be used to obtain results that are more accurate than those generated by the other methods tested.