Gilchrist MJ , Christensen MB , Bronchain O , Brunet F , Chesneau A , Fenger U , Geach TJ , Ironfield HV , Kaya F , Kricha S , Lea R , Massé K , Néant I , Paillard E , Parain K , Perron M , Sinzelle L , Souopgui J , Thuret R , Ymlahi-Ouazzani Q , Pollet N .

Wellcome Trust

	Figure 1. The manual mark-up annotation tool in use. The annotator is presented with the in situ image to annotate and the appropriate schematic diagram (which they may override). The task is to identify the regions of in situ stain on the image and transfer them as faithfully as possible to the schematic, making mental allowance for any rotation, inversion or more complex distortion required. In situ images are shown on the top row, annotations on the schematic diagram are shown on the bottom row; red represents strong staining, orange is weak staining. On the left, an image of a gastrula stage 10 embryo is shown, viewed from the vegetal pole with the blastopore side up. In the middle, a stage 15 neurula is shown viewed from the anterior region, dorsal side up. On the right is shown a stage 22 early tailbud, viewed from the side; dorsal is to the top, but in this case the embryo has been photographed unconventionally with anterior to the right, and the annotator has had to flip the annotation 180° on a vertical axis. Schematic diagrams: copyright 1994 from “Normal Table of Xenopus laevis” by Nieuwkoop and Faber. Reproduced by permission of Garland Science/Taylor & Francis, LLC.
	Figure 2. Results from the comparison annotation test set. All 21 annotators were presented with a test set of 16 images to annotate over the course of the jamboree, to get an informal assessment of annotation confidence levels. The images spanned a wide range of development stages. A: The in situ images and the annotation heat maps for two of the consistently annotated test images, where red shows agreement over all annotators. B,C: The in situ images and the heat maps for two less consistently annotated images, where annotators have not agreed about staging. The test in situ images are shown side by side with the corresponding schematic diagrams. A: The gastrula stage 10 embryo is shown, viewed from the vegetal pole with the blastopore side up; the neurula stage 17 embryo is shown as a dorsal view, anterior to the left. B,C: Tail bud stage embryos are shown as lateral views, dorsal to the top and anterior to the right. Annotation is based on cells with dimensions 1/40th the height of the schematic, and in these heat maps the annotation density of each cell is the sum of the individual image annotations performed on each schematic. Schematic diagrams: copyright 1994 from “Normal Table of Xenopus laevis” by Nieuwkoop and Faber. Reproduced by permission of Garland Science/Taylor & Francis, LLC.Download figure to PowerPoint
	Figure 3. Annotation heat maps from stages 19–22. The full set of annotations is available as a set of heat map images based on the schematic diagrams for each stage. Here, we see part of that set, from stages 19–22, clearly showing both where the greatest numbers of annotated images are to be found, and also the approximate distribution of gene expression over the embryo. The schematics are generally oriented anterior to the left and dorsal to the top. N is the number of images annotated on each schematic. Users of the database may click on a heat map to access the underlying annotated images. Annotation is based on cells with dimensions 1/40th the height of the schematic, and in these heat maps the annotation density of each cell is the sum of the individual image annotations performed on each schematic. Schematic diagrams (anterior/dorsal and stage 22 lateral): copyright 1994 from “Normal Table of Xenopus laevis” by Nieuwkoop and Faber. Reproduced by permission of Garland Science/Taylor & Francis, LLC.Download figure to PowerPoint
	Figure 4. Image retrieval with a user-defined expression pattern. The annotated images may be searched with a user defined expression pattern, using the annotation tool to define the pattern of interest. A,B: Two example queries are shown: a narrow central stripe on a neurula stage 13 embryo, displayed from a posterior-dorsal view, anterior to the top (A), and two paraxial stripes on the same view (B). The results obtained are ranked according to the similarity between the image annotation and the query (score on the top), and additional information is shown, such as gene or clone name, and protein sequence similarities for identification purposes. Note the clear discrimination between the two closely spaced patterns in these examples. Schematic diagram: copyright 1994 from “Normal Table of Xenopus laevis” by Nieuwkoop and Faber. Reproduced by permission of Garland Science/Taylor & Francis, LLC.Download figure to PowerPoint
	Figure 5. Image retrieval with an expression pattern from another in situ image. The annotated images may be searched with the expression pattern from another in situ image, although the user remains unaware of the actual annotation data involved. The database uses the expression pattern recorded for the selected image to search against all the expression patterns for that stage and view, ranking the results by the degree of similarity. Here, annotated expression in the somites from an image of a stage 30 embryo is used to instigate a search.Download figure to PowerPoint