Accurate single-sequence prediction of protein intrinsic disorder by an ensemble of deep recurrent and convolutional architectures
File version
Accepted Manuscript (AM)
Author(s)
Paliwal, Kuldip
Zhou, Yaoqi
Griffith University Author(s)
Primary Supervisor
Other Supervisors
Editor(s)
Date
Size
File type(s)
Location
License
Abstract
Recognizing the widespread existence of intrinsically disordered regions in proteins spurred the development of computational techniques for their detection. All existing techniques can be classified into methods relying on single-sequence information and those relying on evolutionary sequence profiles generated from multiple-sequence alignments. The methods based on sequence profiles are, in general, more accurate because the presence or absence of conserved amino acid residues in a protein sequence provides important information on the structural and functional roles of the residues. However, the wide applicability of profile-based techniques is limited by time-consuming calculation of sequence profiles. Here we demonstrate that the performance gap between profile-based techniques and single-sequence methods can be reduced by using an ensemble of deep recurrent and convolutional neural networks that allow whole-sequence learning. In particular, the single-sequence method (called SPOT-Disorder-Single) is more accurate than SPOT-Disorder (a profile-based method) for proteins with few homologous sequences and comparable for proteins in predicting long-disordered regions. The method performance is robust across four independent test sets with different amounts of short- and long-disordered regions. SPOT-Disorder-Single is available as a Web server and as a standalone program at http://sparks-lab.org/jack/server/SPOT-Disorder-Single.
Journal Title
Journal of Chemical Information and Modeling
Conference Title
Book Title
Edition
Volume
58
Issue
11
Thesis Type
Degree Program
School
Publisher link
Patent number
Funder(s)
Grant identifier(s)
Rights Statement
Rights Statement
This document is the Postprint: Accepted Manuscript version of a Published Work that appeared in final form in Journal of Chemical Information and Modeling, copyright 2018 American Chemical Society after peer review and technical editing by the publisher. To access the final edited and published work see https://doi.org/10.1021/acs.jcim.8b00636
Item Access Status
Note
Access the data
Related item(s)
Subject
Medicinal and biomolecular chemistry
Theoretical and computational chemistry
Theoretical and computational chemistry not elsewhere classified