Prediction of bacterial surface exposed proteins Aleksandr Barinov1, Valentin Loux2, Pierre Nicolas2, Philippe Langella3, Emmanuelle Maguin1, and Maarten van de Guchte1. INRA, 1Génétique Microbienne, 2Mathématique Informatique et Génome, 3Ecologie et Physiologie du Système Digestif, 78352 Jouy en Josas Cedex, France Proteins that are exposed at the bacterial cell surface can mediate physical interactions with the biotic and abiotic environment and are important in bacteria-host interactions like adhesion to eukaryotic cells, pathogenicity and immunomodulation. Existing in silico methods predict the global localization of proteins in bacteria, omitting the potential surface exposition of membrane proteins. We therefore developed a prediction scheme with the particular aim of identification of potentially surface exposed proteins from Gram-positive bacteria. This scheme integrates information of the wellestablished algorithms HMMsearch, LipoP, SignalP, and TMMOD. Dataflow through the scheme is directed by logical decisions and the structure of the scheme allows easy updating through the integration of additional data analysis modules or customized parameter settings. The performance of the prediction scheme was evaluated using experimental results that have been reported for the surface proteome of Streptoccoccus pyogenes. The analysis scheme was then applied to lactobacilli of the acidophilus group. This group of lactobacilli is of particular interest as it contains commensal species from the human GI tract, several of which are used as probiotics, as well as the dairy bacterium Lactobacillus delbrueckii ssp. bulgaricus. The comparative analysis of these lactobacilli revealed qualitative and quantitative differences with respect to potentially surface exposed and secreted proteins. Using the results of the SurfG+ analysis, a number of surface exposed and secreted proteins were selected for the further studies. The corresponding genes are expressed in Lactococcus lactis, and resulting recombinant strains are tested in vitro for immunomodulation and adhesion to GI tract epithelial cells. The future application of this method to GI tract metagenome data should facilitate the identification of proteins that may be involved in host-commensal interactions.