You are not logged in. (Sign In)
PI: Lei Wang (Contact, NU), Jose Luis Ambite (USC), Steven Potkin (UCI), Jessica Turner (MRN), David Keator (UCI)
Summary: Large-scale data sharing and integration is needed to further the state-of-the-art schizophrenia research, but is presently not possible due to practical limitations in the way in which data are being shared. We propose a data mediation and integration resource to overcome these limitations in a low-cost manner and deliver a web portal to interact with the federated databases.
The schizophrenia research community has invested substantial resources to develop methods to collect, manage and share neuroimaging data, along with other meta-data such as clinical, behavioral, cognitive and genetics data. The exploration and analysis of multi-site, multi-dimensional, multi-modal data has improved our understanding of the relationships among abnormalities of brain circuitry, brain function and genetic variability in schizophrenia, as demonstrated by efforts from multi-site consortiums such as the Functional Biomedical Informatics Research Network (FBIRN) and the Mind Research Network (MRN) Clinical Imaging Consortium (MCIC). However, to draw meaningful conclusions about these complex measures, data from large samples are needed, far more than what would be possible at any individual site. Presently, such large-scale data analysis and discovery in schizophrenia research using data from multiple sources would depend upon the effort of separately obtaining data from different places. Practical challenges related to data sharing due to cost of duplication of data, disparate database architecture and querying are preventing many research teams from being able to participate in state-of-the-art neuroscience research.
We propose to create a SchizConnect Mediator as a data mediation and integration platform for establishing a true federation of schizophrenia neuroimaging-related databases to support hypothesis generation and testing, and to deliver a web portal, SchizConnect, to interact with the federated databases. In this approach, information from disparate, heterogeneous databases are queried and integrated in a uniform, semantically-consistent structured manner. To a user or a client program the system appears as a single (virtual) database with a uniform schema/model of the domain, but the data remains at the repositories and under the control of the data providers. We will first test and validate this approach on three large data sets and extend this federation to other data sets in the final year.
By bringing information from disparate neuroimaging repositories into a common integrated system, we will provide resources for investigators to conduct unique investigations not addressable at any single site. By integrating existing data sources that are native to their own institutions and heterogeneous in the terminology and modeling constructs, we will enable more schizophrenia researchers to share their already-collected existing data with no changes to their existing data repositories, thus greatly promoting discovery related to the mechanisms underlying schizophrenia, a key NIMH strategic objective.