PredictProtein server

Reinhard Schneider schneider at embl-heidelberg.de
Mon Mar 27 14:04:10 EST 1995



================================================================================
                     THE PredictProtein SERVER (MaxHom / PHD)
================================================================================

                             WHAT IS PredictProtein?

An automatic service for the  prediction  of aspects of  protein structure.  You
send an  amino acid sequence  and   PredictProtein  returns a  multiple sequence
alignment,  and a prediction of the secondary structure,  residue solvent acces-
sibility,  and helical transmembrane regions.   You can submit your sequence ei-
ther by electronic mail, or interactively from World Wide Web.

                           HOW DOES PredictProtein WORK?

First, the sequence database  (currently only SWISSPROT)  is scanned for similar
sequences.  Second, a multiple sequence alignment is performed by a weighted dy-
namic programming method (MaxHom, [1]).  Third, the generated multiple alignment
is used as input for the predictions.
The predictions are produced by profile based neural network systems.   The fol-
lowing levels of prediction accuracy have been evaluated in cross-validation ex-
periments: secondary structure prediction (PHDsec, [2,3]),  expected three-state
(helix, strand,rest) overall accurracy >72% for water-soluble globular proteins;
solvent accessibility (PHDacc, [4]),  expected correlation between  observed and
predicted relative accessibility > 0.5; transmembrane helices (PHDhtm, [5]), ex-
pected overall two-state accuracy (transmembrane, non-transmembrane) > 95%.

                             HOW TO USE PredictProtein?

Using email (internet):
	Send sequences (or for further information the word 'help') to:
		PredictProtein at EMBL-Heidelberg.DE
Using World Wide Web (WWW):
 	Home page:   
                http://www.embl-heidelberg.de/predictprotein/predictprotein.html

                                   REFERENCES

[1]  Sander & Schneider (1991) Proteins 9, 56-68.
[2]  Rost & Sander (1993) J. Mol. Biol. 232, 584-599.
[3]  Rost & Sander (1994) Proteins 19, 55-72.
[4]  Rost & Sander (1994) Proteins 20, 216-226.
[5]  Rost, Casadio, Fariselli & Sander (1995) Protein Science 4, in press.

--------------------------------------------------------------------------------
            Burkhard Rost, Reinhard Schneider, Antoine de Daruvar & Chris Sander 
                     Protein Design Group, European Molecular Biology Laboratory
                                                              Heidelberg, Europe
--------------------------------------------------------------------------------



More information about the Bionews mailing list