Update of signal peptide prediction server:
Sigfind2 - Signal Peptide Prediction Server (Human)
at
http://www.stepc.gr/~synaptic/synaptic.html
(C)opyright 2001 by Synaptic Ltd.
Contact: M. Reczko, reczko at web.de
This software (SIGFIND2) predicts signal peptides
at the start of protein sequences or
searches open reading frames with a potential signal peptide
coded in nucleotide sequences.
NEW: Bidirectional recurrent neural networks (BRNNs) are used for
prediction.
It is trained on the human protein data used for the SIGNALP system
described in
H.Nielsen, J.Engelbrecht, S.Brunak, and G.von Heijne:
"Identification of prokaryotic and eukaryotic
signal peptides and prediction of their cleavage sites"
Protein Engineering, vol. 10 no. 1 pp. 1-6, 1997
The SIGNALP data is derived from A.Bairoch and B.Boeckmann:
"The SWISS-PROT protein sequence data bank: current status",
Nucleic Acids Res. 22:3578-3580 (1994).
Using the same fivefold crossvalidation as SIGNALP,
the 5 networks of SIGFIND2 (average Mathews correlation coefficient
0.99)
perform better than SIGNALP (average Mathews correlation coefficient
0.96).
The predictions of the 5 networks are combined into a jury decision.
The BRNN algorithm is described in "Bidirectional Dynamics
for Protein Secondary Structure Prediction"
P. Baldi, S. Brunak, P. Frasconi, G. Pollastri and G. Soda,
in R. Sun and L. Giles, editors,
"Sequence Learning: Paradigms, Algorithms, and Applications",
Springer Verlag, 2000.
SIGFIND2 can be used for non-commercial evaluation at
http://www.stepc.gr/~synaptic/sigfind.html
Run Sigfind2 on your own PC
You can get the current version of Sigfind to run on your own PC.
Sigfind and its Web-Interface are available for x86-PCs
running Linux (trademark of Linus Torvalds)
or MS-Windows (trademark of Microsoft Corp.).
Send an email to mailto:synaptic_ltd at yahoo.de to get
information about the license fee.
Martin Reczko
Synaptic Ltd
