Full Case Study (.pdf)

Executive Summary Only (.pdf)

Collaboration in biomedical research

The biomedical research community is turning to collaborative research as a model for maximizing progress on major diseases. Collaborative research involves multiple teams working together to solve a problem by sharing their research plans, reagents, specimens, and preliminary data. Researchers working toward an HIV/AIDS vaccine are a prime example of this approach. Progress on an effective vaccine has been stalled by the elusiveness of the virus. In recognition of this challenge, leading research and funding organizations including the National Institutes of Health and the Bill and Melinda Gates Foundation formed an alliance dedicated to accelerating the development of a preventive HIV vaccine. One of the core strategies of this alliance is to promote collaboration by creating more efficient, faster ways for researchers to share successes and failures and avoid duplication of efforts.

To implement their collaboration strategy, NIH funded the Center for HIV/AIDS Vaccine Immunology (CHAVI) and the Gates Foundation funded the Collaboration for AIDS Vaccine Discovery (CAVD). CHAVI and CAVD are research consortia whose members comprise more than 110 institutions in 25 countries. Investigators within the member institutions apply new scientific knowledge and a diversity of cutting-edge research techniques to create and evaluate novel vaccine candidates. CHAVI and CAVD also run many different types of pre-clinical studies, including several large-scale, multi-site studies of people at risk of contracting the HIV virus. These studies collect specimens and observational data on thousands of participants and make these resources available to other consortia members.

Demands on information systems

CHAVI and CAVD researchers are expected to share their project plans and experimental data with other consortium members, as a condition of their funding. This level of coordination clearly requires some support from the computing systems used by the collaborators. The broad reach of the Internet makes such large scale collaboration conceivable and increasingly achievable. It is safe to assume today that every scientific researcher has the computing resources to connect to the Internet. Beyond that basic level of commonality, however, there will be wide variation in the systems used, software installed, and networking protocols available. Because of this variation, the systems for sharing documents in a large consortium must be different from the combination of email and file servers that are commonly used for such purposes within a smaller group in a single organization.

Sharing experimental data among collaborators is even more difficult than sharing project documents. For example, each of the many different experimental techniques used by CHAVI and CAVD generate different types of result data files. In some cases these files are too large to be analyzed by desktop computers and tools. In other cases the files are so numerous that they become difficult to track and compare. Ultimately, the experimental data must be analyzed and validated and then integrated with observational data and communicated in ways that allow other consortia members to find and interpret the results. As with documents, the systems that work for scientific data sharing within a single organization often do not scale well to a large-scale collaboration. Lastly, information systems that support collaboration demand agility in their design and implementation. Building a system that can adapt easily to the requirements of different collaborators on a project is important to getting them to use the system effectively. The system (and the people behind it) must be flexible in order to adapt to changes in requirements and suggestions for improvements from collaborators. The urgency of the research goals for large-scale collaborations and the uncertainty of how they will be achieved also drive tight implementation schedules and a flexible and iterative development approach.

LabKey Server: Software solutions for collaborative science

The Statistical Center for HIV/AIDS Research and Prevention (SCHARP) was selected to provide both data management and statistical analysis expertise to the CHAVI and CAVD consortia. SCHARP is part of the Vaccine and Infectious Disease Institute of the Fred Hutchinson Cancer Research Center based in Seattle, Washington. SCHARP formed a team to manage the project of building and maintaining the system, nicknamed “Atlas”.

SCHARP needed information systems for the CHAVI and CAVD projects that would be flexible enough to adapt to rapidly evolving research protocols and diagnostic instruments. Commercial software packages could be used for specific tasks, but the overall system required a comprehensive, custom solution. SCHARP chose to partner with LabKey Software to design and build that solution. LabKey brought development expertise in high-throughput web applications, a skill set that complemented SCHARP’s strength in designing and managing biomedical research studies. The open-source LabKey Server platform provided the foundation for a customized solution, a foundation that was already proven in production use on other large research projects. The Atlas system was built as a set of plug-in modules on the LabKey Server platform.

Today, the Atlas system, based on LabKey Server, provides three essential services to both consortia:

• Secure Collaboration: Atlas includes a set of web-based collaboration tools ideally suited for sharing documents and short notes. These support several common web formats, including wiki pages, message boards, and issue trackers. All of these collaboration tools are thoroughly integrated with a security model that is both thorough and simple to understand. Atlas security integrates with SCHARP’s internal authentication system and provides secure user account management for external collaborators. This makes administration of groups and permissions straightforward.
• Assay Data Management: Atlas takes a comprehensive approach to managing data from diagnostic tests and laboratory instruments by providing customizable tools to import, analyze, validate, and integrate these datasets in a central SQL database. The system includes an assay designer that allows SCHARP staff to specify the file data format for results to be loaded into the database, as well as speedy data entry forms for capturing additional information to collect with the data file. Once the data is in the database, SCHARP researchers can employ a range of web-based tools to filter, sort, compare, search, visualize and compute statistics on the data. In the future, CHAVI and CAVD will use these tools in combination with external tools to ensure the quality of shared data.
• Study Management: Atlas supports the design of a study as an information framework for managing the data collected in the course of a research investigation on a set of subjects over time. The study framework keeps track of the participants in the investigation (the subjects) as well as the specimens to be analyzed and the time points that are relevant to the analysis. The study framework includes functions for data validation and navigation, and specimen request tracking. CHAVI, for example, runs several large observational studies consisting of hundreds of human participants around the world. The participants follow a schedule of regular clinic visits at which specimens are collected and clinical measurements are recorded. The remote clinics record their measurements on paper forms and send them via DataFax to SCHARP. Specimens are divided into vials and kept in central storage facilities until a researcher requests one or more vials to run a diagnostic assay. Assay data and metadata are linked into the same study framework. The integration of all of these datasets enables the study framework to provide built-in views that show summaries per time point, and allow SCHARP statisticians drill down by cohort and participant to find trends and anomalies.