Nitya Thakkar, a junior computer science major at Brown University, used machine and deep learning methods to understand the spatially coordinated mechanisms of tumor survival in Classical Hodgkin’s Lymphoma
Classical Hodgkin’s Lymphoma (CHL) is a cancer of the lymphatic system. While many cancers have homogenous or less diverse tumor microenvironments, CHL is unique since it has a very diverse tumor microenvironment that is composed of only 1-5% tumor cells and the remainder are nonmalignant immune cells. This summer, I have grown so much as a researcher and individual through BSRP. By talking with other scientists at the Broad and developing my research skills through my project, I have realized that I do want to pursue a Ph.D. and continue to tackle new and innovative questions. I have really appreciated the opportunity to learn from my accomplished peers and Broad researchers this summer, and am so excited for what I will do next using the skills I have learned this summer!Previous studies have found that these tumor cells are dependent on the tumor microenvironment for survival and to evolve their mechanisms of immune evasion. However, not much is known about what specific cells and genes may aid in their survival. Therefore, the goal of this project is to develop new computational methods and apply them to spatial transcriptomic data to understand how cell-specific gene expressions change in different regions of the tissue.
First, we created a neural network to identify the types of cells in the spatial transcriptomic data. We trained the model on single cell reference data where the gene expression inputs and cell type outputs were well defined. We then tested the model on the spatial transcriptomic data, where the model predicted the cell types based on gene expression values. Next, we aimed to determine if there is any relationship between the frequency of a certain cell type and the distance to the nearest tumor cell. We found that dendritic cells, which represent a major cell type in the tumor immune microenvironment, increase in frequency as distance to the nearest tumor cell decreases. We will next investigate how cell states change in tumor dense areas. We anticipate that these methods will help us learn more about cellular interactions and gene expression changes in the tumor microenvironment, which will allow us to develop more targeted therapies in the future.
Project: Determining the spatially coordinated mechanisms of immune evasion in Classical Hodgkin’s Lymphoma
Mentors: Neriman Tokcan, Caroline Uhler, Todd Golub