Justus Florian Gräf - PhD Scholarship 2024

Project summary:
Genetic regulation of blood plasma traits and their impact on complex cardiovascular diseases

Investigating the genetic regulation of circulating proteins and metabolites in the blood can help us to understand the causes and consequences of complex cardiovascular conditions such as heart failure, coronary artery disease and hypertension. Identifying complex patterns in the genome and interactions of genetic variants with dietary patterns can advance risk stratification in cardiovascular diseases (CVD).

Project Title

Genetic regulation of blood plasma traits and their impact on complex cardiovascular diseases

Background

Many cardiovascular diseases (CVD) are complex and cannot often be attributed to a single gene but are influenced by a mixture of genetic and environmental factors. This complexity makes it difficult to identify genetic predispositions in patients. Circulating proteins and metabolites in the blood can serve as biomarkers for CVD, and in some cases, altered levels can be causal for the development of CVD. A major challenge is to identify how genetic variants lead to disease, and understanding the complex genetic regulation of biomarkers can help bridge the gap between the genome and the disease phenotype.

Aim

The project will focus on the following aims:

Train Deep Learning models to identify complex genetic patterns and gene-diet interactions that influence the blood plasma proteome and metabolome.

Transfer of pre-trained deep learning models to predict individual-level blood plasma protein and metabolite levels in Danish cohorts to advance risk stratification in CVD

Methods

Using Deep Learning algorithms, we will integrate genetics, proteomics, metabolomics and electronic health record data from large population-based cohorts such as the UK Biobank (~500K individuals). The resulting models can be further analyzed to identify complex effects such as genetic epistasis, dominance or gene-environment effects that influence the blood plasma proteome the metabolome and ultimately the risk to develop CVD. Lastly, we can use the pre-trained models from the UK Biobank on other cohorts to predict protein and metabolite levels, which can help to advance risk stratification in CVD.

Preliminary results

So far, we trained and analyzed Deep Learning models on genetics and proteomics data from the UK Biobank and identified complex genetic effects for a small group of proteins. The results of this first part of the project are published as a preprint (https://doi.org/10.1101/2024.07.04.24309942).