High-throughput microscopy (HTM) is a powerful analytical and diagnostic tool for cancer research. Visualizing a large number of individual cells, HTM generates vast spatiotemporal information on cellular structures and behavioral dynamics under a multitude of physiological conditions[2,3] computationally evaluating the whole aspect of such data could produce unbiased, systematic representations of biological processes[4,5], often unveiling hidden features in bulk or small population measurements. HTM has driven major breakthroughs in cancer research, particularly in screening anti-cancer drugs and identifying resistance mechanisms. Despite HTM’s benefit and potential, technical and practical issues hinder the wide-use of the platform in routine laboratory settings. Conventional microscopes achieve either a large field of-view (FOV) or high resolution, but not both simultaneously. Most HTM systems thus are constructed by combining high-magnification microscopes with mechanically scanning stages. This scheme increase the system complexity and cost, while limiting image acquisition speed. Also, analyzing massive amount of imaging data from HTM is a daunting and labor-intensive. Although recent advances in imaging software have significantly reduced this burden[7-9], further improvements are needed in analyzing multi-dimensional images (e.g., time-lapse movies), and automating feature extraction.This project will advance a new HTM approach, termed deep learning microscopy (DLM). DLM will adopt cutting-edge developments in computational optics and deep learning, addressing technical challenges in the current HTM. Computational optics can overcome fundamental limits of conventional optics by exploiting the power of digital imaging and fast computation. This proposal will specifically use the computational illumination approach, known as Fourier ptychographic microscopy (FPM)[10-12]. FPM collects low resolution image sequences while changing the position of a point-light source; each image has spatially-shifted spectrum information in Fourier space, where the amount of shift depends on the Illumination angle. By numerically stitching image sequences, the Fourier space, including high frequency components, can be reconstructed, which enables FPM to achieve high spatial resolution even with low magnification lenses. Deep learning (DL) is making impressive progress in its ability to derive information from large data sets, to the point where such techniques can outperform human analyses for many data sets[13,14] DL is particularly potent in discovering intricate, hidden structure in high dimensional data sets, and has been adopted for solving quantum physics problem. This proposal will leverage DL power to analyze large images from FPM. We hypothesize that deep learning can extract multidimensional features of individual cells, revealing uncharacterized dynamic cellular processes.We will develop a DLM platform with the capacities for i) high-resolution, wide FOV imaging; ii) molecular profiling on individual cells; and iii) automated imaging analyses (molecular expression, morphology, migration, propagation) on large numbers of single cells. The DLM will transcend the current microscopy. It will have unprecedented analytical power: imaging large number of individual cells at high spatial resolution, and automatically extracting multitude of cellular features. As such, we envision that the DLM will be a transformative tool for cancer research. Potential applications include better monitoring anticancer drug responses, analyzing cellular heterogeneity in large section of tissues, and prospectively detecting cellular fate under various physiological perturbations.