Deep learning for protein structure prediction

Dnr:

SNIC 2017/11-7

Type:

SNAC Large

Principal Investigator:

Arne Elofsson

Affiliation:

Stockholms universitet

Start Date:

2017-07-01

End Date:

2018-07-01

Primary Classification:

10203: Bioinformatik (beräkningsbiologi) (tillämpningar under 10610)

Secondary Classification:

10610: Bioinformatik och systembiologi (metodutveckling under 10203)

Tertiary Classification:

10601: Strukturbiologi

Webpage:

http://bioinfo.se/

Allocation

Abstract

A complete structural knowledge of all proteins and their interactions in a cellular compartment would provide an unprecedented detailed picture of the compartment. Such a model would make it possible not only to ask what functions and regulation could occur but also what could not. A few years ago it would be completely unrealistic to even obtain the structure of all proteins and even more unrealistic to obtain details about their interactions. But there has recently been a revolution in structure prediction for both individual proteins and complexes. The basis for this is the development of contact predictions methods using direct coupling information. Here, we apply for computational resources to continue the development of our tools using deep learning approaches and to apply them for proteomic level structure prediction projects. In summary we develop methods that combine deep learning methods with molecular simulations techniques in a unique way to obtain 3D-structure of proteins and complexes.