Developing pleiotropy analysis and inference testing for discovery of functional networks from high dimensional datasets (Prof. Christoffer Nellaker)
PROJECT TITLE
Developing pleiotropy analysis and inference testing for discovery of functional networks from high dimensional datasets.
SUPERVISORS
-
Christoffer Nellaker
-
Amanda Chong
DESCRIPTION OF PROJECT
The project focuses on developing methods and tools to analyse high dimensional multi-omic, genetic, and deep phenotyping datasets. The goal is to create tools for discovery of functional linkage maps and thereby enable discovery of mechanistic biology and novel drug repurposing targets.
We have developed a proof-of-principle framework for utilizing pleiotropic effects on emergent processes to infer if different markers influence related functional pathways. We call this tool pleiotropy analysis and inference testing (PLAIT). A successfully candidate would take this PLAIT forward in two interlinked primary streams of work:
- Method improvement
- Biological discovery
Method development:
PLAIT stands at the intersection of machine learning and statistical genetics. Currently PLAIT implements neural networks and traditional statistical models, but these will not scale to imaging and deep high dimensional phenotyping modalities. We will explore vision-based attention and transformer modelling, and along-side fine-tuning of foundation models to accelerate and empower learning.
PLAIT outputs a directed graph of putative functional linkages between markers. We will investigate graph network analyses of PLAIT outputs to quantify transitive leakage of functional linkage through the network. We will explore methods for causative inference through PLAIT derived functional spaces.
Biological discovery:
We will explore PLAIT on different biological data modalities and disease settings. In line with implementation of modelling approaches into the PLAIT codebase we will deploy this on UKB and GTEx datasets to investigate the functional pathways associated with metabolomic, proteomic, anthropometric and morphological phenotypes. Primarily these investigations will be carried out in collaboration with groups with interest and expertise in specific disease settings.
TRAINING OPPORTUNITIES
In the course of this project the successful candidate will be embedded within a team covering statistical genetics and machine learning. There will be opportunities to engage in formal training courses to build skills in areas the candidate would benefit from a more general introduction.
Through the course of a DPhil project with us the successful candidate will be given opportunities to learn and practice the skills required for rigorous, innovative and fundable scientific research.
Funding Information
Unfunded, however, the Deep Medicine group will assist the DPhil candidate in securing funding.
HOW TO APPLY
To apply for this research degree, please click here.