An integrative, data-driven, and computational approach to uncovering dynamic mechanisms of early viral infection

Project Summary Cutting-edge technologies are generating large datasets across biological processes, including those following viral infection and host responses. However, lack of computational tools that can extract meaningful insights, and lack of ability to integrate information across different model systems and data modalities, are roadblocks to

deriving biological and mechanistic understanding of these processes. The recent rise of devastating viruses including SARS family viruses reveals that a deeper, basic mechanistic understanding of viral infection is still lacking. Specifically, new insights into early viral infection (asymptomatic replication phase) and early-responding

genes that govern infection and disease outcome are critical for understanding progression of infection and host responses. During my postdoctoral research, I developed several widely-used algorithms for biomedical machine learning and single-cell data analysis, and applied these to a broad range of biological systems, including

infectious disease. Here, I propose to develop a completely new approach that is founded in cross-modal computational analysis and can be applied to dynamic processes across living systems. In this proposal, the method will be trained upon and applied to uncovering virus infection dynamics. By leveraging single-cell

technologies, combinatorial CRISPR perturbation, and advanced machine learning, this new approach will learn the gene regulatory logic that governs infection. By spanning model systems, I will extract information that can be derived more cleanly from in-vitro systems, such as early infection timepoints. Through cross-integration of

these data with in-vivo data from mouse models we will bring the precision questions that can be asked in human organoids together with the physiological environment of animal models, powering our ability to derive relevant insights into gene networks underlying a complex, dynamic process. I will build a single-cell atlas of virus infection

and train a machine learning algorithm to obtain a predictive model of infection dynamics. By also integrating data from single-cell combinatorial CRISPR perturbation, I will infer causal gene networks as well as synergistic gene interactions that govern infection dynamics. This combination of advanced machine learning methods,

large-scale single-cell analysis, and gene perturbation data will allow discovery of the drivers of infection, signatures of both susceptibility and protection, and gene networks that can ultimately be targeted for therapeutic intervention. Synergistic gene interactions will open up future paths to potentially more effective, specific, and

even combinatorial therapies. The innovative coupling of computational methods and deep data collection to extract information, particularly during early infection phases, has the potential to fundamentally change our understanding of viral infections, as well as provide a framework that can be applied to a broad range of biological

processes and diseases to obtain deep mechanistic understanding.