The genus mycobacterium includes two important human pathogens: Mycobacterium tuberculosis and Mycobacterium leprae.  However, their virulence factors and the mechanisms of bacterial pathogenesis in leprosy and tuberculosis are still ill defined.  In a previous report our group has identified 5 genes coding for putative secreted proteases of M. leprae that are expressed in vivo during the course of human infection.  Bacterial proteases are potential virulence factors and are excellent candidates for drawing new drugs for controlling infection.  In the genomic era methods for large scale analysis are necessary not only for correct protein annotation, but also for focusing on specific protein categories or predicting their level of interaction with the host.  In this study we used bioinformatic tools to analyze the set of proteases predicted in mycobacterial annotated genomes and search for proteases initially not assigned as such.  We compared M. leprae protease set with proteases present in M. tuberculosis H37Rv and Mycobacterium bovis AF2122 genomes, which coding sequences (CDS) have been predicted and annotated by the Pasteur Institute (http://genolist.pasteur.fr).  Through this approach 39, 51 and 50 proteases were identified in M. leprae, M. tuberculosis and M. bovis, respectively.  M. leprae proteases were then used as templates to identify the homologues present in Mycobacterium avium paratuberculosis str. K10 genome, which its complete sequence and all gaps has been closed by University of Minnesota (www.cbc.umn.edu/researchprojects/AGAC/mptb/mptbhome.html).  An identical set of proteases was found in M. paratuberculosis except for one enzyme (ML2613).  This analysis revealed other interesting findings such as the identification of a unique protease present in M. tuberculosis but not in M. bovis and M. leprae.  Currently molecular modeling analysis is underway to predict the three dimensional structures of selected enzymes for potential therapeutic application.  Supported by CNPq and FAPERJ.