Public archiving and data integration in the era of multi-modal imaging

Consider a scenario in which one could image a whole organism, and then progressively zoom in to the level of organs, tissue, cells, molecules and finally atoms, at each step revealing more detail. Although the prospect of scaling from the level of organisms to atoms may seem far-fetched, there now do exist imaging techniques to cover almost all of these imaging scales. There is a growing body of biomedical research where information from different imaging techniques at different scales are combined, i.e., multi-modal imaging, to piece together a more comprehensive understanding of a problem than can be obtained from one technique alone. However the wider integration of imaging data is hampered by the lack of public accessibility to this data and commonly used formats for representing the information required to bring the different modalities into register. To address all scales of imaging at once would be extremely challenging given the enormous scope of the problem, the disparate user communities and the resources required.
In the field of molecular structural biology, which involves the study of the 3D structure of molecules, the concept of public archiving of structures is well established with archives such as the PDB (atomic model data) and EMDB (3D volumes derived from electron microscopy (EM) experiments including electron tomography) enjoying widespread community support. Our strategy has therefore been to first consider the integration between the cellular and molecular imaging scales in consultation with the relevant communities. We organized two 'expert' workshops inviting key specialists and community representatives which resulted in concrete recommendations including - exploring public archiving of large EM image datasets, the development of visualisation tools that provide an integrated, problem-centric view of structures piecing together information from the different imaging modalities, the imaging modalities which should be targeted first - 3D scanning electron microscopy (3DSEM), soft X-ray tomography (SXT) and correlative light and electron microscopy (CLEM) , and the development of a format to facilitate data integration. In an on-going MRC-funded project (grant MR/L007835; MOL2CELL) we have successfully developed EMPIAR - a public archive for raw EM image data, which is capable of handling data-sets in the terabyte range. We are developing a web-based visualisation tool (Volume browser) that provides an openly accessible integrated view of 3D biological data at the cellular and molecular scales derived from the EMDB and PDB archives, and a format that supports one form of integration between these scales. In the proposed project we want to build upon the work started in the MOL2CELL project and a) extend archiving in EMPIAR to include full support for the proposed imaging modalities, b) extend format support to represent the information required to bring the different modalities into register and use this to improve and extend the Volume browser, and c) develop software and data pipelines from major microscopy centres to facilitate data deposition to EMDB and EMPIAR and to improve the quality and quantity of the information collected.
While the scope of the project is limited to the molecular and cellular imaging scales, the outcomes are potentially applicable to other imaging modalities. The inclusion of the new imaging modalities into EMPIAR can serve as one model for how public archiving can be organized and established, the format developed can be used for data integration between other imaging modalities and the Volume browser for integrated visualisation with other imaging scales. Furthermore developers of resources for multi-modal imaging data stand to gain from open access to the software developed in the project and the data in EMPIAR.

The use of imaging modalities spanning the transition from the molecular to the cellular imaging scales such as electron tomography (ET), 3D scanning electron microscopy (3DSEM), soft X-ray tomography (SXT), and correlative light and electron microscopy (CLEM) in biomedical research is growing rapidly. At present there is no established public archiving for this data except for ET data, which is archived in EMDB. In consultations with the EM and cellular imaging communities it is clear that extending public archiving to include these imaging modalities and improving data integration between different scales of imaging would confer enormous benefits with regards to the potential reuse of the data, enabling for instance the problem-centric visualisation of structural data where information from the cellular and molecular scales are combined. In the context on an on-going MRC funded project (grant MR/L007835; MOL2CELL) we have developed EMPIAR - a pilot archive for raw 2D image data related to EMDB entries, and are in the process of developing a web-based Volume browser that will provide integrated visualisation of EMDB and PDB entries where the link between the molecular and cellular scales is provided by segmentation information. In the proposed project we want build on the momentum and experience from the MOL2CELL project to a) extend EMPIAR to support 3D Scanning Electron Microscopy (3DSEM), Soft X-ray Tomography (SXT) and Correlative Light and Electron Microscopy (CLEM) data, b) develop standards for correlative imaging data and for relating sub-tomogram averages to tomograms, and extend the Volume browser to accommodate CLEM visualisation as well as overlays of sub-tomogram averages on tomograms, and c) develop data-harvesting and submission pipelines to EMDB and EMPIAR to facilitate deposition and to improve the quality and quantity of experimental information collected.

Public archiving of 3D Scanning Electron Microscopy (3DSEM), Soft X-ray Tomography (SXT) and Correlative Light and Electron Microscopy (CLEM) data
Public archiving of these modalities will benefit end users in the structural biology and cellular imaging communities who want to reuse it, e.g., for data mining or validation. EMPIAR already serves as a useful source of test data for methods developers and this will also apply to developers working with these imaging modalities. It will lessen the archiving burden on individual labs. It will help journals improve the quality and transparency of publications related to these imaging modalities. It can also serve as a model for other bio-imaging communities for how image archiving can be set-up, managed and organized.

Visualisation of and standards for correlative imaging and for relating sub-tomogram averages to tomograms
The transformation format and related libraries will benefit software developers in the tomography and correlative imaging fields by reducing the burden on them to handle conversions between different packages.
The extensions to the Volume browser will enable the integrated visualisation of CLEM, 3DSEM and SXT data in EMPIAR with data from EMDB and PDB. It will thus be a unique publicly accessible tool providing exceptional added value to the data stored in these archives. The problem-centric views to structural data that it will provide will create opportunities for data-mining and knowledge discovery by the wider biomedical community. For teachers and students of biology and biochemistry the resource will provide an evidence-based visual exposition of many key biological systems and processes. For the pharmaceutical industry the resource will open new avenues to visually understanding how potential drug targets behave in the context of the cell environment. The extensive linkage of information will also mean that the resource will rank highly in common search engines like Google when searching for more general biological and medical terms used by the general public. It will thus provide a valuable entry point for general users to develop a better visual appreciation for the basis of biology and medicine on the micro scale.

Data-harvesting and submission pipelines to EMDB and EMPIAR
These developments will substantially simplify the deposition process and reduce the time it takes thereby enabling the microscopy centres and depositors to cope with the expected rise in data volumes in the future. We will also work with the Cambridge Pharmaceutical Cryo-EM Research Consortium to better understand the requirements of industry and tailor the data-harvesting and pipelining software so that it could more broadly be used by other industry related microscopy centres. These developments will also lead to an improvement in the quality and quantity of the information being collected about the experiment which will benefit the end users of the data, primarily the structural biology and cellular imaging communities.