GPI Lipid Anchor Project

The learning set

Contact:
Birgit Eisenhaber (IMP/Austria)
Peer Bork (MDC/EMBL)
Frank Eisenhaber (IMP/Austria)

SWISSNEW(27th of Jan, 1999)-part of the learning set (total 9 entries)

SWISSPROT(rel.36)-part of the learning set (total 160 entries)

Comments to the selection of entries for the learning set

Origin of information on GPI-modification sites

Annotated proteins with known GPI-modification carry the keyword "GPI-anchor" and the token "FT LIPID" in SWISS-PROT.

This set of 177 entries contains information on 126 metazoan sequences, 40 protozoan sequences, one viral sequence and 10 fungal sequences. The omega-sites for ACES_TORMA and ACES_TORCA have been edited in accordance with new literature data (Bucht and Hjalmarsson, BBA, 1996, 1292, 223-232).

Four protozoan sequences attracted attention as a result of their deviating propeptide length and other strange sequence properties. In all four cases, the omega-site annotated in the database was not verified experimentally. Two entries (GP46_LEIAM, GP63_LEIGU) were deleted due to their extreme propeptide length (>31 residues) and the absence of an obviously reasonable alternative omega-site. In two entries with extremely short propeptide (<16 residues), the omega-site was edited in accordance with homology considerations and/or an expert sequence property analysis. The final learning set for protozoa consists of 38 sequences.

Out of the original set of 126 entries, 6 were deleted due to extreme propeptide length and the absence of an alternative omega-site. For another 21 sequences, the site has been edited in accordance with homology considerations and/or an expert sequence property analysis. In all 27 cases, the omega-site was not validated with an adequate experimental method. Thus, the final learning set contained 120 sequences.

Last modified: 12th June 2000