Bioinformatics-2017 DeepLoc：prediction of protein subcellular localization using deep learning

Paper: DeepLoc：prediction of protein subcellular localization using deep learning

Project website: http://www.cbs.dtu.dk/services/DeepLoc

DeepLoc: Prediction of Subcellular Localization

Abstract

Prediction of eukaryotic protein subcellular localization is a widely studied problem in bioinformatics because of its relevance to proteomics research. However, most predictors depend on annotations of homologs drawn from knowledge databases. For novel proteins lacking annotated homologs, and to predict the effects of sequence variants, it is desirable to have methods that infer protein properties from sequence information alone.

We present a prediction model and the DeepLoc dataset.

Introduction

Proteins carry out a broad and diverse range of functions in the distinct compartments of eukaryotic cells. A protein’s function depends on the compartment or organelle in which it resides, because that context provides the physiological setting for its activity. Aberrant subcellular localization can alter the functions a protein displays and contribute to the pathogenesis of many human diseases, including metabolic, cardiovascular, and neurodegenerative disorders, as well as cancer.

Prior work has relied on annotated proteins or on sequences with closely related annotated homologs.

Materials and methods

Results

<HR align=left color=#987cb9 SIZE=1>