Abstract
Metagenomic analysis has become a powerful tool to analyze bacterial communities in environmental samples. However, the detection of a specific bacterial species using metagenomic analysis remains difficult due to false positive detections of sequences shared between different bacterial species. In this study, 16S rRNA amplicon and shotgun metagenomic analyses were conducted on samples collected along a stream and ponds in the campus of Hokkaido University. We compared different database search methods for bacterial detection by focusing on Legionella pneumophila. In this study, we used L. pneumophila-specific nested PCR as a gold standard to evaluate the results of the metagenomic analysis. Comparison with the results from L. pneumophila-specific nested PCR indicated that a blastn search of shotgun reads against the NCBI-NT database led to false positive results and had problems with specificity. We also found that a blastn search of shotgun reads against a database of the catalase-peroxidase (katB) gene detected L. pneumophila with the highest area under the receiver operating characteristic curve among the tested search methods; indicating that a blastn search against the katB gene database had better diagnostic ability than searches against other databases. Our results suggest that sequence searches targeting long genes specifically associated with the bacterial species of interest is a prerequisite to detecting the bacterial species in environmental samples using metagenomic analyses.