Polymerase chain reaction (PCR) has been widely used in the detection of mycobacteria. PCR is particularly applicable where the relevant pathogens may be difficult to culture or are in low abundance. The overall sensitivity of PCR assay may however be substantially reduced due to a vast excess of non-target DNA and inhibitory substances in the sample. pPN14 plasmids containing IS900 were digested with SmaI/SacI to remove the optimal M. ptb specific PCR 413 target site from the 5' region and religated. Primers p8 5'-TbGTGGCGTTTTCCTTCGGTG-3' and p21 5'-GbCGCTCGAGTAGCCGCGTTC-3' were then used to amplify 5'-biotinylated 513bp capture probes from the 3' end of IS900 where homology with IS902 is greatest. In the test, 50mcl of buffer containing an excess of 5'-b capture probe was added to 450ml as sample DNA extract, boiled for 5 min, incubated overnight 65°C, cooled 25°C and 15mcl of a suspension of streptavidin-coated M280 Dynabeads added. After 2h the beads were captured magnetically and the supernatant discarded. PCR IS900 and IS902 was carried out directly on the beads. Titration experiments demonstrate the ability of this system to detect 2-20 M. ptb in a ml of buffer. 100mg fecal samples from 10 bovines suspected of Johne's disease and of shedding M. ptb in low abundance were boiled in 1ml NaI Geneclean buffer for 20 min and the DNA extracted on glass beads. Direct IS900 and IS902 PCR was negative on all samples. Mycobacterial DNA was then captured as described and PCR repeated. 6 animals tested strongly positive for M. ptb and one for M. avium subsp silvaticum (M. avs). Hybridization capture may contribute to the simple automatable detection of M. ptb and M. avs in clinical and environmental samples.