LF-TIMING: Large-scale label-free profiling of cell-cell interactions - 7. Project Summary/Abstract There is a compelling need for technologies that can quantitatively profile cell-cell interactions on a large scale, especially the interactions of immune cells with tumor cells and quantify detailed measurements of crucial events including cell-cell contact patterns, contact frequencies, contact durations, cell death, interaction-related movement patterns especially pursuit, avoidance, arrest, movement failures, and target cell killing behaviors, cell division, and others. The TIMING™ (Time-lapse Imaging Microscopy In Nanowell Grids) assay is a powerful and versatile high-throughput in vitro assay that meets this need. In its current form, TIMING assay analysis software relies on multi-channel fluorescent imaging to identify cell type and location. This imposes an inherent limitation, as even the best available high-throughput time-lapse imaging instrumentation has 4 – 5 fluorescent channels of which 2 – 3 channels must be devoted to labeling effector and target cells, leaving only 1 – 2 channels for the customer to incorporate antibodies or probes of investigational interest. There is a compelling need to free up fluorescence channels so the customer can incorporate 3 – 4 investigational probes concurrently. This will greatly enhance the value of the TIMING assay for the customer by allowing them to study the localization and dynamics of multiple investigational probes concurrently in their spatiotemporal context. Achieving this goal requires computer vision methods capable of accurately and reliably analyzing label-free phase-contrast video recordings to detect, segment, and classify cells, track their movements, detect key events like cell death and mitosis, and profile cell-cell interactions while coping with focus loss. This will not only free up multiple fluorescent channels but also reduce the phototoxicity, enabling cell co-cultures to be imaged over longer durations and at higher frame rates. Finally, there is a need for the video array analysis to be completed while the cells are still live (~ 2 hours maximum), to identify crucial cells of interest (e.g., exceptionally motile serial killers) for robotic retrieval and downstream processing, including clonal expansion and transcriptomic profiling. The goal of this project is to develop LF-TIMING (Label Free TIMING), an integrated computer vision system that meets the above-mentioned needs, leveraging advances in deep neural network-based cell segmentation, tracking, classification, and focus restoration methods.
