Project Reference: ICS27-SCI-ST2

Project title: From genetic variation to immune cell function: combining cell-type-specific deep learning with experimental validation

Project outline

A fundamental goal of functional genetics is to understand how non-coding genetic variation controls gene regulation in a cell-type and state specific way. Immune cells are an ideal system for addressing this question as population-scale studies have discovered thousands of regulatory variants influencing immune traits. Interpreting these signals has proven difficult, as most are non-coding and act in highly specific cellular contexts. Deep learning models can predict functional consequences of genetic variation, but remain limited in their ability to capture this cell-type and state-specificity, leaving an important gap in linking genome sequence with gene regulation and cellular function.

This project aims to bridge the gap by building a deep learning model that predicts cell-type-specific effects of regulatory variants, and by experimentally validating its predictions. Within the project, the PhD candidate can choose a computational or an experimental emphasis. The computational aims are to 1) fine-tune a pre-existing foundation model for improved cell-type-specific predictions by integrating multiple large-scale single-cell sequencing datasets from immune cells, and 2) benchmark the fine-tuned model against existing models and experimental data across various contexts, including variant classes and cell types. The experimental aims are to 1) identify regulatory mechanisms of immune-associated variants by assessing them with the fine-tuned model, and 2) validate the predicted regulatory mechanisms using perturbation models, and confirm associated changes in cell composition and state using flow cytometry and immunohistochemistry.

The Teichmann lab is ideal for this project as it is at the forefront of single-cell analytics and has done pioneering work in unravelling immune cell function and regulation. It has collected massive single-cell datasets, has strong expertise in deep learning modelling, and offers extensive experimental and computational resources.

Principal Supervisor

Professor Sarah Teichmann

Email address: sat1003@cam.ac.uk

Department / Partner Institute: Cambridge Stem Cell Institute

Website: http://www.teichlab.org/

Industrial Partner

OtoImmune Ltd

BBSRC strategic theme

Understanding the rules of life