| dc.description.abstract | At the heart of any human disease is an imbalance between normal and aberrant physiological processes— a disproportion between hypo-immunity and hyperimmunity—a lack of homeostasis. In many cases, a more comprehensive understanding of the molecular basis underlying disease progression and therapeutic failure is still required to devise new strategies for improving patient outcomes. Technological advancements in biomedical research, especially in single-cell omics (e.g. single-cell RNA sequencing, single-cell spatial profiling) have given us unprecedented power to decipher the intricate cellular and molecular features that maintain—or disrupt—this balance. However, validating the causality of these features remains a huge challenge, as the wealth of data often results in a considerable number of hypotheses to test. In this thesis, I explore applications of single-cell genomics tools to understand cellular features associated with disease, with a particular focus on tuberculosis (TB). I then present a potential solution for performing phenotypic screens at scale. In the first part, I applied single-cell RNA sequencing and analysis to human lung samples from a TB-endemic region in South Africa. Using contrastive analysis, I identify key cell populations that are differentially abundant between TB-diseased and TB-negative lung including several neutrophils, macrophages, and fibroblasts subsets. I discovered a de novo gene program highly enriched in the MMP1+CXCL5+ Fibroblast that correlates with TB burden in a non-human primates (NHP) granuloma dataset, supporting the importance of this subset in TB. In a collaborative effort, we validate that this MMP1+CXCL5+ Fibroblast localizes to TB granuloma on independent TB-diseased lung tissues using immunohistochemistry assays and recapitulate the induction of this population from lung-derived fibroblast through in vitro stimulation experiment with M.tb. I further report a SPP1+ macrophage population that is enhanced in TB diseased lungs through single-cell analysis. Moreover, I identified a prominent cross talk between SPP1+ macrophages and fibroblasts in TB diseased lung that mimics similar observations in cancer and fibrosis, supporting an important role for this axis in TB. These distinctive cell populations could serve as potential targets for novel host-directed therapies in tuberculosis. In the second part, I developed a method to compress small molecule phenotypic screens by designing randomized drug pools with replicates of distinct candidates across different drug pools. Our team demonstrated that linear regression models can be applied to computationally deconvolute the individual hits, enabling the identification of top effectors for downstream validation. We benchmarked and demonstrated the efficacy of this approach in a cost-effective imaging platform and then moved into applications on pancreatic ductal adenocarcinoma (PDAC), where we discovered a new perturbation response signature to IL-4/IL-13 with prognostic value for patient survival. We also showcased the utility of this tool on understanding immunomodulation effects in heterogenous mixtures of primary blood cells. Together, this thesis describes novel cellular features important to TB in human lungs, offering new insights that complement existing knowledge from animal models. It also presents a bold, yet effective strategy to scale up phenotypic screen across different biological systems, providing a much-needed solution that bridges the translational gap between human disease and experimental model. | |
