BIBLIOS

Document type
Conference papers

Document subtype
Full paper

Title
Selenoprotein discovery using neural networks

Participants in the publication
Alves, M. (Author)
Lourenço, C. (Author)
Dep. Informática

Summary
Identifying SECIS sequences in the genome is an essential step in Selenoprotein discovery. We use a competition neural network in a pre-processing stage to encode potential SECIS sequences, and a multilayer perceptron trained with backpropagation to predict SECIS. We thus propose a new approach within SECIS studies by combining more classical bioinformatics techniques with the learning and generalization capabilities of neural networks. Selenoproteins are thought to be responsible for most biomedical effects of dietary selenium and are essential to mammals. They play an important role in cancer prevention, immune function, aging, male reproduction, and other physiological and pathophysiological processes. Their prediction and study are of undeniable value. The ability of artificial neural networks to be fault tolerant, to adapt easily to new environments without specialized programming and to handle fuzzy, probabilistic, noisy or inconsistent data allowed us to provide an alternative to stricter pattern matching and score based methods. Our results compare well with the ones obtained with previously available methods.

Editor(s)
Nuno Guimarães and Pedro Isaías


Event
IADIS International Conference on Applied Computing 2005, February 22-25

Publication Identifiers

Organizers
IADIS

Publisher
IADIS