By Topic

Use of artificial neural networks in sequential models of membrane proteins of the Swiss-Prot database for the detection of transmembrane segments

Sign In

Cookies must be enabled to login.After enabling cookies , please use refresh or reload or ctrl+f5 on the browser for the login options.

Formats Non-Member Member
$33 $13
Learn how you can qualify for the best price for this item!
Become an IEEE Member or Subscribe to
IEEE Xplore for exclusive pricing!
close button

puzzle piece

IEEE membership options for an individual and IEEE Xplore subscriptions for an organization offer the most affordable access to essential journal articles, conference papers, standards, eBooks, and eLearning courses.

Learn more about:

IEEE membership

IEEE Xplore subscriptions

2 Author(s)
A. C. Kalpakas ; Technical University of Crete, Department of Electronic and Computer Engineering, Polytechnioupolis, Chania 73100, Greece ; M. A. Christodoulou

Geneticists try to find the hidden truth behind the genomes which contain the blueprint for all parts of life's machinery. Next challenge will be corresponding DNA to various types of proteins; thus deriving meaningful knowledge for the understanding of biological systems. Proteins are capable of explaining human evolution while others, such as membrane proteins, can reveal the developing mechanisms of diseases, such as muscle disease, blindness, diabetes, arthritis, and cancers. Transmembrane segment topology is crucial for protein's folding into space and understanding its role inside and outside of a cell. For that reason, a feed-forward neural network is constructed using the back-propagation algorithm, aiming at the prediction of transmembrane segments (helices) in membrane proteins out of a single amino acid sequence. More than three hundred human and non-human proteins (extracted from the Swiss-Prot database) are used to train the network. Variable input and output sequence lengths urged the implementation of a sliding-window technique. Several configurations are tested and optimal parameters, such as weights, learning rate and window sizes are elaborated. The system is evaluated based on its ability to successfully predict the topology of new, unknown membrane proteins: Performance reaches 94.23% and test MSE is held down to 5.77%. Sensitivity factor is equal to 93.20% and specialty factor goes up to 96.17%.

Published in:

Control & Automation, 2007. MED '07. Mediterranean Conference on

Date of Conference:

27-29 June 2007