eMouseAtlas as a community resource for data sharing
eMouseAtlas has been a major MRC-funded programme for 16 years and is entering its fourth phase. Phase 1 developed the key atlas design, spatiotemporal database functionality, and basic image-processing tools. Phase 2 introduced an Editorial Office (EO) and converted the prototype to a functioning system with selected data. Phase 3 lifted the database from a relatively small system to a major international resource, unique in scope and content and embedded as a key resource for development and molecular genetics research. In this fourth phase, we propose to extend the database and web browser interfaces to make sure the whole research community can contribute to the eMouseAtlas as an integrating framework. Uniquely, we have begun to add atlas material from other sources as a step to widen community involvement in the production of a definitive resource. Thus far, additional atlas models include whole embryo (Dr T. Mohun, NIMR); forelimb and hind-limb models (Delaurier et al. 2008); kidney (EuReGene); Caltech MRI whole-embryo delineated models (Dr. S. Ruffins), and five to six somite embryos cultured for 24 h (Cajal et al. 2012). Furthermore, we propose to offer a portal for data sharing amongst the developmental biology community. The short-term goal for this project is to deliver a practical community resource that provides sharing and publication of 2D/3D embryo images. Whereas existing social networking services such as Facebook allow open sharing of 2D images, there is no facility in place to share, view, and publish 3D images using the World Wide Web, and none that can provide the scientific annotation and metadata tracking. eMouseAtlas provides interactive access to very-high-volume 3D images via a tiled pan-and-zoom-style interface, similar to the approach utilised by GoogleMaps. In addition to being very efficient, WlzIIP3D can be embedded in a standard tiled-image server to provide dynamic tile generation directly from virtual sections cut through an image volume. We have already demonstrated this technology for very large single images (up to 150 Gb), but this will extend to the multi-Tb images now being generated in biomedical research. We wish to extend the use of IIP3D technology by allowing researchers to upload, share, view, and publish 3D images using an interactive web-based application. The initial implementation will allow user data upload, sharing, and publication of 3D data. Critical in developing this interactive web-based app is the sharing of views and work in progress under the control of the user. The server technology that has been developed at eMouseAtlas, in particular, the ability to perform online queries of a high-volume 3D image archive, renders the eMouseAtlas resource perfectly suited as a long-term repository for large-scale 3D biological images such as those provided by OPT and MRI. Many research projects have a condition of funding that data are made publicly available at the end of the project and it is anticipated that the eMAP repository will provide a structured and managed mechanism for researchers to satisfy this requirement.
eMouseAtlas and Kaufman’s atlas of mouse development
eMouseAtlas is negotiating a licensing agreement from the publishing house Elsevier to deliver an online version of plates from The Atlas of Mouse Development by Matt Kaufman (1992). In collaboration with Elsevier, we are developing an online interface that uses the IIP3D server to deliver high-resolution colour images of the original histological sections used in Kaufman’s atlas (Fig. 6). The original sections have been digitised to generate >1,000 images that correspond to the plates used in the atlas. These plates are annotated using Kaufman’s original terms plus the corresponding eMAP ontology terms enabling direct links to eMouseAtlas and the exploration of partonomic relationships, gene-expression patterns, and “wiki”-based definitions and descriptions associated with each term.
eMouseAtlas and spatial transcriptomics
EMAGE has comprehensive transcriptome data in the form of sparse section series. We have developed a trained colour-segmentation model to separate signal (in situ hybridization) from histology and background in the EurExpress data set, and we map this data in two ways. The first is a rapid-pass pseudo-whole-mount approach to provide an overview pattern for query and analysis. This is already in use in EMAGE and about 15,000 patterns from serial section data (EurExpress) have been spatially mapped. The extended approach maps this sparse section data into the full 3D space of the model embryo and interpolate the data to enable 3D query and analysis. This combined with the automatic signal extraction provides an automatic process to map the data in 3D. This is currently being tested on the EurExpress and EmbryoExpress data sets (19,000 transcripts).
In addition, we plan to apply a transcriptomic tomography technique to the embryo to deliver a low-resolution but complete spatial transcriptome using next-generation sequencing technologies. Drs. Okamura-Oho and Yokota of the RIKEN Wako Institute have developed a tomographic technique to capture the full-expression map of the murine brain (Okamura-Oho et al. unpublished). The technique is based on tissue sectioning with block-face imaging to allow reconstruction in conjunction with tissue collection and expression analysis (using standard microarray chips). In collaboration with Drs. Okamura-Oho and Yokota, we propose to pilot the technique for whole-mouse embryos. We will use Next Gen RNA-Seq to capture the full transcriptome of the 14.5-dpc embryo and to enable cross-validation of expression profiles with EurExpress in situ hybridisation data. To generate a 3D map, embryos will be serially sectioned in each of three orthogonal planes: frontal, sagittal, and transverse. If this strategy is successful, then we will develop a larger-scale proposal to apply this technique across a full range of stages from 9.5 to 17.5 dpc to deliver low-resolution but transcriptionally complete maps of the mouse embryo.
In addition to delivering the complete spatial transcriptome at key stages of mouse development, we propose to develop online tools that would allow analysis of spatial data and to compare this with more conventional text-based annotation. It is noteworthy that a tool for comparative analysis is already in place as part of the Argudas project (McLeod et al. 2012). The Argudas tool evaluates in situ gene-expression patterns by comparing text annotation associated with EMAGE and GXD entries. An extension of this tool that evaluated spatial annotation would be useful in delivering comparisons between spatial data and text-based annotation.