PhD Candidate University of Wisconsin - Madison Madison, Wisconsin
Body of Abstract: A seedling’s ability to control growth at the systemic and cellular level is essential for proper development, where plant-level changes in size and shape occur by regulating expansion at the cellular level. Quantifying this growth can help to elucidate the mechanisms underlying its control; however most analyses do not measure the dynamic behavior of the growing material that produces these higher-level changes. This requires frame-by-frame tracking of small elements along the seedling, traditionally done by manually selecting elements on the image or by external application before imaging. These are low in throughput is cumbersome in methodology. Specifically, the process of segmenting the seedling can be hindered by unusual conformations, such as when the cotyledon is blocking the stem from view. To address the need for more efficient methods, we developed HypoQuantyl, a kinematics tool that uses a novel machine learning approach to eliminate many of the issues with traditional segmentation. We trained a system of machine learning algorithms on images with manually-drawn contours around a seedling’s growth region - or hypocotyl, which were tasked to place contours when shown new images. This system reliably segmented seedlings typically difficult for traditional methods. We then performed frame-by-frame tracking on patches of texture along the midline. From the motion of these patches we measured its Relative Elemental Growth Rate, which describes that seedling’s growth kinematics. This pipeline is executable on cloud computing servers, allowing for high-throughput analysis. Using this tool, we identified changes in the growth profile of Arabidopsis thaliana mutants deficient in the blue-light sensing photoreceptor CRY1, which exhibits an escape from an initial light-induced growth inhibition. This observation provides insight into the localization of CRY1 and furthers our understanding of how photoreceptors contribute to seedling de-etiolation.