Image-derived, Three-dimensional Generative Models of Cellular Organization

摘要: Given the importance of subcellular location to protein function, computational simulations cell behaviors will ultimately require ability model distributions proteins within organelles and other structures. Towards this end, statistical learning methods have previously been used build models sets two-dimensional microscope images, where each set contains multiple images for a single pattern. The learned from not only represents pattern but also captures variation in that cell. consist sub-models nuclear shape, organelle size shape distribution relative boundaries, allow synthesis with expectation they are drawn same underlying as train them. Here we extend generative approach three dimensions using similar framework, permitting locations be described more accurately. Models different patterns can combined yield synthetic multi-channel image containing many desired, something is difficult obtain by direct imaging than few proteins. In addition, parameters represent compact interpretable way communicating descriptive features, may particularly effective automated identification changes organization caused perturbagens.