Deep learning for high-throuput data

Dnr:

SNIC 2017/3-83

Type:

SNAC Small

Principal Investigator:

Mika Gustafsson

Affiliation:

Linköpings universitet

Start Date:

2017-08-31

End Date:

2018-09-01

Primary Classification:

10203: Bioinformatik (beräkningsbiologi) (tillämpningar under 10610)

Webpage:

Allocation

Abstract

A common problem with associating gene variants to complex disease associated traits is power and multiple testing. We have used gene networks to incorporate functional relevance among gene variants, but has not lead to decision support systems. In this project we aim to integrate network information with deep learning utilizing genomics in millions of gene variants and 1000's of patients and integrating transcriptomics and gene networks. For this purpose we will need to use TENSORFLOW and the GPU architecture available at Kebnekaise. The project is funded by grants from SSF and VR.