Our immune system can help us prevent or slow cancer development. Human CD4⁺ T cells play critical roles in regulating our immune responses to fight cancer. Upon encountering a pathogen, naïve CD4⁺ T cells differentiate into different T helper (Th) cells to perform diverse immune-modulatory functions. Variability in this differentiation process is associated with variable responses to cancer immunotherapy. While several genes necessary for differentiation have been identified, researchers lack a comprehensive map and a predictive model of the larger gene regulatory network (GRN) controlling this process. Dr. Zhu [Connie and Bob Lurie Fellow] plans to combine functional genomics with mathematical modeling to systematically map and model the human CD4+ T cell differentiation GRN and use the GRN model to predict and control the differentiation process. His work promises to provide a quantitative understanding of the CD4+ T cell differentiation process and open up new strategies for safer and more effective cell-based cancer therapy. Dr. Zhu received his PhD from the California Institute of Technology, Pasadena and his BS from Hong Kong University of Science and Technology, Hong Kong.

Dr. Zweig [Sijbrandij Foundation Quantitative Biology Fellow] is modeling how gene regulation changes across time and space using advanced geometry, deep learning models, and CRISPR gene-editing technology. With these tools, it is possible to learn how RNA expression evolves in a cell over time and, ideally, predict evolution in regions susceptible to cancerous mutation(s). This work has the potential to apply to many types of cancer but is especially applicable for acute myeloid leukemia (AML). This is in part because there are known “precursor” states—when cells are at higher risk of becoming cancerous—where understanding gene regulation is most impactful, and in part because a common element of AML treatment is a stem cell transplant, which may attack host cells inconsistently in different regions of the same organ.

The project models temporal gene dynamics via stochastic differential equations, parameterizing the drift vector field with linear networks or shallow neural networks to guarantee provable identifiability, trained via differentiation through the adjoint method. Spatial interactions between spot clusters are characterized with graph neural networks, symmetric functions parameterizing summary statistics, and self-attention modules on latent gene embeddings. The latent embeddings are also defined through a variational autoencoder integrating RNA with other modalities.