Deep integrative omic auto-encoder analysis
Title: Deep integrative omic auto-encoder analysis
SNIC Project: Berzelius-2021-26
Project Type: LiU Berzelius
Principal Investigator: Mika Gustafsson <mika.gustafsson@liu.se>
Affiliation: Linköpings universitet
Duration: 2021-08-18 – 2021-12-01
Classification: 10203
Homepage: https://gitlab.com/Gustafsson-lab
Keywords:

Abstract

We recently created a deep auto-encoder by analyzing ~70,000 RNA-seq samples and developed a new reverse-training method to identify functionally similar and potential disease-causing genes (Dwivedi et al Nature Communications 2020). The last 1,5 years we have analyzed DNA genomics and epigenomics (methylation) using the largest repositories (UKBiobank ~500,000 samples). Our goal is to create a deep data-driven multi-omics autoencoder that also includes prior biological knowledge and can be re-used to minimally represent the core data in smaller clinical studies for multiple purposes and be used for identification of disease-causing mechanisms.