This study evaluated conventional methods, GLC and three molecular tests, including 16S rRNA sequencing, for the identification of mycobacteria, and the experiences of the authors with the integration of these methods into a diagnostic clinical laboratory were also recorded. Of 1067 clinical isolates of mycobacteria identified by conventional tests, 365 were tested by Accuprobe hybridization assays and PCRs specific for Mycobacterium tuberculosis (MTB) complex or Mycobacterium avium complex (MAC), 202 were tested by 16S rRNA sequencing, and 142 were tested by GLC. Three runs of all tests were performed on a weekly basis. The identifications for 209 MTB complex and 118 MAC isolates obtained by species-specific PCR were in complete agreement with AccuProbe hybridization and conventional test results. The 16S rRNA sequence-based identification, at a similarity of > or =99 %, for 132 of 142 isolates was concordant with the identifications made by the biochemical methods, and for 134 isolates was concordant with the identifications made by GLC at species, group or complex level. 16S rRNA sequencing resulted in fewer incorrectly identified or unidentified organisms than GLC or conventional tests. For the slowly growing non-tuberculous mycobacteria, the mean turnaround times for identification were 4-5 days for 16S rRNA sequencing, 14-29 days for GLC and 22-23 days for conventional methods. Considering the large proportion of some species among clinical isolates, a strategy of initial screening with species-specific PCR (or AccuProbe assays) for the MTB complex and MAC, followed by direct sequencing of the strains that yield negative results, should make 16S rRNA sequencing more affordable for routine application in diagnostic laboratories.