Pasquier, C. and Hamodrakas, S.J.
Department of Cell Biology and Biophysics,
Faculty of Biology, University of Athens
Panepistimiopolis, Kouponia, Athens 15701, Greece
PRED-TMR is a method that predicts transmembrane domains in proteins
using solely information contained in the sequence itself. The algorithm
refines a standard hydrophobicity analysis with a detection of potential
termini ("edges", starts and ends) of transmembrane regions. This allows
both to discard highly hydrophobic regions not delimited by clear start
and end configurations and to confirm putative transmembrane segments not
distinguishable by their hydrophobic composition.
We have now extended this application with a pre-processing stage represented
by an artificial neural network which is able to discriminate with a high accuracy
transmembrane proteins from soluble or fibrous ones.
Applied on several test sets of transmembrane proteins, the system gives a perfect
prediction rating of 100% by classifying all the sequences in the transmembrane class.
Applied on 995 non-transmembrane protein extracted from the PDBSELECT database, the
neural network predicts falsely 23 of them to be transmembrane (97.7% of correct assignment).
The neural network is described in :
Pasquier C, Hamodrakas SJ:
An hierarchical artificial neural network system for the classification
of transmembrane proteins. Protein Eng 1999 Aug;12(8):631-4
