On the other hand, methods for structure prediction have improved tremendously in applicability, speed and confidence, which can be used to bridge the wide gap between sequence and structure . Here we describe an effort to annotate the structures of the TB proteome. Annotation has been achieved through a pipeline that integrates a number of different computational approaches. Obtaining the models at a genome scale provides one of the first opportunities to view the structural profile of the proteins in the organism and understand the cellular functioning in terms of structural scaffolds that facilitate the underlying molecular recognition events. To obtain a structural LDN193189 proteome of M. tuberculosis H37Rv, an integrated structural annotation pipeline was developed. Each model is annotated with metrics rating its quality, high and low confidence regions in the model and a range of features suggestive of its function. High-quality molecular models for 2511 proteins were derived from Modbase and an additional 54 were derived through remote homology detection methods. Put together with 312 crystal structures from PDB, structural models for 2877 proteins are thus available. Functional annotation procedures as applied to the 2877 models involved detection of conserved residues in the protein family, location of ligand or DNA binding site , similarity with enzyme active sites, and finally possible ligand associations. In all, 1728 ligand associations have been found, providing functional clues. Given the genome-wide nature of this study, two types of coverage are considered. First, the coverage of the genome or the number of different proteins Iressa modeled and second, the coverage of each protein or the number of residues in the individual polypeptides that could be modeled was analyzed. 2877 proteins, which is ,70% of the proteome, have been modeled . 1427 of these showed complete length coverage, , whereas 2233 models had length coverage of at least 50%. Typically coverage is lower for proteins in cell wall processes or insertion sequences categories. The proteins showing least coverage belonged to PE/ PPE and other membrane proteins, since in these, quite often only a domain could be modeled. The procedures for modeling protein structures are now well established , most often resulting in models correct to an RMSD of ,2 A�� . As a validation exercise, crystal structures of proteins that are solved have been compared with the corresponding models obtained through Modpipe workflow by excluding the crystal structures in the template search step.