Immune cells, such as T helper cells, innate lymphoid cells (ILCs), and macrophages, are traditionally classified into discrete subsets based on their expression of specific transcription factors and effector cytokines. The lens of single-cell RNA sequencing (scRNA-seq) has revealed a more complex spectrum of immune cell states and functional plasticity across states. Standard, clustering-based analyses often have limited power to model such phenomena. Instead, we represent the continuum of cell states using latent factorization techniques, which allow for cells and genes to have a gradient of membership in different programs. These methods improve the power and interpretability of downstream analyses in many of our applications. Thus, they play a role in several ongoing projects, including the following:

• Studying how the thymic development and T-cell receptor (TCR) signaling of human γδ T cells relate to their immune function

• Adapting methods to learn interpretable, interconnected gene programs

• Developing methods for estimating RNA velocity and analyzing spatial transcriptomics

• Integrating histology and ‘omics data to discover better cancer biomarkers

Who is involved: Hope Anderson, Frank Gao, Hanna Hieromnimon, Joseph Sifakis, Ruxandra Tonea