ABSTRACT Imaging-based single cell transcriptomics technologies create a single molecule resolution map of near complete transcriptome in native tissues, unlocking the long-standing dream of comprehending the spatial organization of molecules and cells in intact tissues. The structural organization of molecules and cells is closely tied to their functional organization, thus, transcriptome-scale RNA imaging would provide invaluable insights into how molecules and cells interact and collectively perform systems-level functions in healthy and diseased tissues. Among different technologies, MERFISH (multiplexed error-robust fluorescence in situ hybridization) occupies a leading position with its high spatial resolution, high detection efficiency, single molecule sensitivity, and high multiplexing capability. However, current technologies are not fast enough to process tissue blocks of any meaningful size, leaving critical questions like 3D tissue profiling, cross-tissue comparisons, and large-scale atlas efforts out of reach. Here, we propose to close this gap by at least an order of magnitude by combining custom biotechnology with modern statistics to build a next-generating imaging-based single cell transcriptomics platform. We will develop experimental techniques and analytical procedures for 1) hyperspectral imaging and 2) computational deconvolution of optically crowded RNA molecules. Few efforts along these directions exist, and no method has proven to be effective. The biggest hurdle is the absence of real experiment-based reference datasets with known ground-truth signals, without which no new methods can be properly validated. For each strategy, we propose to generate a high-quality MERFISH reference dataset as well as develop new statistical models and inference procedures to recover the true signals. Our proposed methods can be integrated with each other and with other approaches to increasing throughput. In long term, we aim to create an in situ single-cell platform that can profile millions of cells in >100mm2 tissue volumes within a day and perform large-scale comparative studies of thick tissue/organ blocks. This will enable multi-tissue analysis, comparative studies of relevant tissue volumes, and large-scale atlas establishment, thereby unlocking new dimensions of human genome research.

NARRATIVE Understanding the structure, function, and secret lives of cells is key to treating and preventing disease, and recent breakthroughs in imaging-based spatial transcriptomics have for the first time made it possible to profile cells in their natural communities within tissues. However, current technologies are not fast enough to process tissue blocks of meaningful sizes, leaving critical questions like 3D tissue profiling, cross-tissue comparisons, and large-scale atlas efforts out of reach. Here, we propose to close this gap by at least an order of magnitude by combining custom biotechnology with modern statistics to build a next-generating imaging-based single cell transcriptomics platform.