A Neural-Network-based system is
trained and tested on a set of well annotated proteins to tackle the problem
of predicting the signal peptide in protein sequences.
The method trained on a set of experimentally derived
signal peptides from Eukaryotes and Prokaryotes, identifies the presence
of the sorting signal and predicts their cleavage sites.
The accuracy in cross-validation is comparable with previously presented programs
reaching the 97%, 97% and 95% for Gram negative, Gram positive and
Eukaryotes, respectively.
(
Click here for the cross-validation sets )
