A0294
Title: Deep learning for biological sequence analysis: The SignalP software
Authors: Konstantinos Tsirigos - Technical University of Denmark (Denmark) [presenting]
Abstract: Deep learning models have been successful in numerous applications such as image, text, and speech recognition. They have become increasingly popular amongst the machine learning tools for bioinformatics during the recent years owing to the availability of greater computational resources, more data, new training algorithms and easy-to-use libraries for implementation. We have used deep learning in a widely-known topic of biological protein sequence analysis and, specifically, the detection of signal peptides in amino-acid sequences. Signal peptides are intrinsic signals for secretion in both eukaryotic and prokaryotic proteins. Since their existence was demonstrated in 1975 by Gnter Blobel (who later received the Nobel Prize for it), there has been a keen interest in the question of how signal peptides actually look and whether they can be predicted from the amino-acid sequence alone. One of the most used methods for making such predictions, SignalP, which has been online since 1996, is now released in its fifth major version, using a deep learning architecture.
