Protein interface prediction using graph convolutional networks
Date
2017
Authors
Fout, Alex, author
Byrd, Jonathon, author
Shariat, Basir, author
Ben-Hur, Asa, author
Journal Title
Journal ISSN
Volume Title
Abstract
Determining the interface between two interacting proteins can help illuminate cellular biology, improve our understanding of disease, and aid pharmaceutical research. Such determination is expensive and time consuming using wet-lab experiments, which has motivated the development of computational methods. Inspired by the success of deep learning in image processing and other application areas, we adapt convolutional neural networks to work with irregularly structured data, such as proteins. We construct a novel pairwise classification architecture which is trained and tested with data from the Docking Benchmark Dataset versions 4.0 and 5.0. This outperforms the existing state-of-the-art prediction method, PAIRpred.
Description
This poster was presented at the 2017 Colorado State University Graduate Student Showcase, 9 November 2017.
Rights Access
Subject
proteins
protein interface
machine learning
neural networks
convolutional neural networks
graph structured data
graphs
deep learning