A sensitive predictor for potential GPI lipid modification sites in fungal protein sequences and its application to genome-wide studies for Aspergillus nidulans, Candida albicans Neurospora crassa, Saccharomyces cerevisiae and Schizosaccharomyces pombe

The fungal transamidase complex that executes glycosylphosphatidylinositol (GPI) lipid anchoring of precursor proteins has overlapping but distinct sequence specificity compared with the animal system. Therefore, a taxon-specific prediction tool for the recognition of the C-terminal signal in fungal sequences is necessary. We have collected a learning set of fungal precursor protein sequences from the literature and fungal proteomes. Although the general four segment scheme of the recognition signal is maintained also in fungal precursors, there are taxon specificities in details. A fungal big-II predictor has been developed for the assessment of query sequence concordance with fungi-specific recognition signal requirements. The sensitivity of this predictor is close to 90'%. The rate of false positive prediction is in the range of 0.1%. The fungal big-H tool successfully predicts the Gas1 mutation series described by C. Nuoffer and co-workers, and recognizes that the human PLAP C terminus is not a target for the fungal transamidase complex. Lists of potentially GPI lipid anchored proteins for five fungal proteornes have been generated and the hits have been functionally classified. The fungal big-II prediction WWW server as well as precursor lists are available at http://mendel. imp.univie.ac.at/gpi /fungi/ gpi-fungi.html (C) 2004 Elsevier Ltd. All rights reserved.