Automated Delineation of Cells and Nuclei and Quantification of Gene Expression in 3D Images of Whole Drosophila Blastoderm Embryos.
1) Life Sciences and Genomics Division, Lawrence Berkeley National Laboratory, Berkeley, CA.
2) Institute for Data Analysis and Visualization, University of California, Davis, CA.
The Berkeley Drosophila Transcription Network Project is developing methods that will allow us to generate a database that records the relative levels of expression of hundreds of genes in each cell of a blastoderm embryo. Here we describe development of a key component of this analysis pipeline: image analysis algorithms that automatically identify the location and extent of all nuclei and their associated cytoplasm in 3D images of embryos and the expression levels of genes in each cell/nucleus. Confocal microscopy is used to acquire 3D images of whole embryos that are fluorescently labeled for the genes of interest and total DNA (see poster by Keränen et al.). Novel image segmentation routines delineate the nuclei based on the DNA stain and estimate the associated cytoplasm by tessellation of the surrounding volume. Intensity attenuation, which increases with imaging depth, is normalized using the DNA image. These methods yield a list of all nuclei, containing their 3D coordinates, morphological features and relative gene expression. Validation of the resulting segmentation and gene expression maps is challenging because this is the first time such comprehensive analysis has been attempted. Novel 3D rendering methods, which allow interactive scoring of segmentation accuracy (see poster by Weber et al.), reveal a small number of errors resulting from the interaction between inherent acquisition anisotropy and embryo geometry. By comparing the local density of nuclei within an embryo and between embryos, we estimate our current technique has a 6% under-segmentation error rate. The accompanying poster by Keränen et al. describes the use of our methods to measure variation in the expression of a set of patterning genes between embryos.
Last modified April 9, 2005.