< Back to previous page

Project

Identifying disease genes and mechanisms through nonlinear fusion of omics data

As different types of omics data (genomic variation, gene expression, proteomics, etc.) keep accumulating at ever increasing speed, integratively analyzing such data to facilitate the experimental identification of which genes and variants are associated or causative for disease, better understand pathogenic mechanisms, and predict potential drug target or therapeutic strategies is also growing challenge. Starting from earlier work on gene and variant prioritization from our lab, and from a variety of methods for nonlinear data fusion (including kernel methods, Bayesian matrix factorization, and deep learning), we will develop an integrative framework that enables the joint modeling of gene, pathway, phenotype, and disease information based on variety of omics data and information sources (such as annotation or text mining of the literature). Methodologically, we will focus on deep learning models of latent representations from incomplete and scarce data in an entity-relation framework that enables the integration of information about multiple types of objects (here, genes, pathways, phenotypes, and diseases). Starting from the modeling of publicly available data, we will focus gradually on specific collaborations with biomedical researchers on several use cases from rare genetic disorders, complex disorders, or cancer to demonstrate the broad relevance of our approach.

Date:8 Sep 2020 →  Today
Keywords:data fusion, genome interpretation, omics
Disciplines:Computational biomodelling and machine learning
Project type:PhD project