Sainsbury Laboratory (SLCU) Yu Group: Decoding the regulatory grammar underlying plant cell identity with AI foundation model
This is a BBSRC Targeted project and applications must be made via the Cambridge Biosciences DTP PhD Programme website. Project Reference: TRG27-SLB-HY1
Supervisor
Dr Haopeng Yu
Project outline
Plants build many different cell types from the same genome. This diversity is achieved largely by switching different combinations of genes on and off in different cells and at different developmental stages, encoded in what is known as a gene regulatory network (GRN). However, constructing GRNs experimentally is challenging because the regulatory grammar is hidden in non-coding DNA with no clear rules, and experimentally testing each gene interaction takes years. As a result, it remains unclear how much of this regulatory grammar is shared across cell types versus rewired to produce distinct cell identities and developmental transitions.
In this project, the student will use state-of-the-art AI techniques to decode the grammar of the gene regulatory networks. It builds on methods already established and running in the Yu lab, including AI agent workflows and the fine-tuning, prediction and explainable-AI toolkit for plant DNA foundation models. Using an AI agent workflow to assemble existing cell-type-specific omics data across development, the student will fine-tune a plant DNA foundation model to predict gene regulation in a cell-type-aware manner, then apply AI interpretability methods to extract the regulatory relationships the model has learned. These will first be benchmarked against mechanistic models of plant development established in the Jönsson group, recovering experimentally confirmed interactions forms an achievable first milestone and establishes that the approach captures real biology. From that validated base, the student will extend the analysis genome-wide, generating new, testable hypotheses about how regulation is rewired as cells change identity or developmental stage, gaining a holistic view of how one genome is read differently to build the many cell types of a plant.
Contact
BBSRC strategic theme
Understanding the rules of life