Automated Annotation of Developmental Stages of Drosophila Embryos by Image Analysis. Jieping Ye^1,2, Lei Yuan^1,2, Qian Sun^1,2, Cheng Pan^1,2, Michael McCutchan¹, Stuart Newfeld³, Sudhir Kumar^1,3. 1) Center for Evolutionary Medicine and Informatics, Biodesign Institute, Arizona State University, Tempe, AZ; 2) Computer Science and Engineering, Arizona State University, Tempe, AZ; 3) School of Life Sciences, Arizona State University, Tempe, AZ.

   Images capturing spatial patterns of Drosophila gene expression are being produced at a higher throughput than ever before. Automated and efficient tools for analyzing these images are a prerequisite for generating biological insights into gene function, interactions and networks. These analyses are the most biologically meaningful when images from a similar time point during development are compared. We present a computational method to automatically annotate the developmental stage of Drosophila embryos displaying gene expression images. The method is based on our observation that image texture changes as embryonic development progresses. Our system is able to accurately determine the development stage of embryos de novo with high accuracy (79%) employing the Campos-Ortega and Hartenstein stage demarcations. The method can also predict ages within the standard developmental stages (e.g., Early versus Late age for a given stage). The within stage information is value in making evaluations about absolute timing of gene expression initiation or alteration. From an analysis of high throughput image data, we found that the Genomewide-Expression-Maps (GEMs) generated using images from embryos in with highly refined stages illuminate global gene activity and transitions better that those employing the standard stages alone. A more precise knowledge of developmental stage also improves the investigator's ability to predict interacting genes when embryonic expression patterns matches are discovered. We will also present results from additional new computational methods that automatically orient, align and annotate expression images with a specific suite of embryo characteristics (e.g., early stage and lateral view).