New prediction server for signal peptides

Martin Reczko martin at stepc.gr
Wed May 23 06:46:39 EST 2001


New prediction server avaliable:

Sigfind - Signal Peptide Prediction Server (Human)
at
http://www.stepc.gr/~synaptic/sigfind.html

(C)opyright 2001 by Martin Reczko (martin at stepc.gr)

This software (SIGFIND) predicts signal peptides at the start of protein
sequences.
A novel neural network learning algorithm is 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 SIGFIND (avgerage Mathews correlation coefficiant
0.98)
perform better than  SIGNALP (avgerage Mathews correlation coefficiant
0.96).

It should be noted that the performances on the test-sets in SIGNALP
are used as a  stop criterion during the training of the neural
networks,
whereas the test-sets in SIGFIND are not used in any way during the
training.

The predictions of the 5 networks are combined into a jury decision.

Martin Reczko
Synaptic Ltd.




More information about the Proteins mailing list