Abstract
Structural analyses of cell are important to bridge molecular functions and phenotype. Recently, various bioimaging technologies are developed and obtained microscopic images are organized as databases. On the other hand, image analyses are depends on human observations and manual measurements in many cases. We are currently developing efficient method for image evaluation of plant cell structures. As a target of analysis, we used confocal images of cytoskeleton and organelles visualized in tobacco BY-2 cells. To evaluate the distribution of images, we tried to utilize clustering analyses. In prior to the clustering, we must develop appropriate distance metrics between images, because standard metrics were lacked in bioimaging analyses. We explored optimizing methods to select the metrics using metadata including annotations. As a result, we developed new clustering algorithm, which can adapt various observation targets. Using this algorithm, we evaluate the cytoskeletal and organellar dynamics during cell cycle progression.
