Presented at:
BECON/BISTIC 2004 SYMPOSIUM - Biomedical Informatics for Clinical Decision Support:
A Vision for the 21st Century, Bethesda, MD, June 21-22, 2004
Appling Standard Data Warehousing Approaches to Biomedical Data
Parrish D, Raab M
The Immune Tolerance Network is a collaborative research effort that solicits, develops, implements and assess clinical strategies and biological assays for the purposes of inducing, maintaining and monitoring tolerance in humans for kidney, liver and islet transplantation, autoimmune diseases and allergy & asthma. The data generated by these trials include clinical endpoints and mechanistic assay data; the ITN is building a data management system that focuses on the curation and integration of numerous and diverse data sources.
Purpose: This presentation will describe our adaptation of a generalized data warehousing strategy which has allowed us to insulate the data repository from the evolving fluid nature of bioassays technologies and techniques by collecting and using metadata to drive the ETL process. Then from this stable repository we can restructure data sets (also driven by the metadata) based biological concepts, assay idiosyncrasies and users questions and deliver views of those data sets through a online analytical processing (OLAP) tool
Methods: The Extract Transformation and Load (ETL) process is accomplished using a modified Data Junction engine which allows for both data source versioning and the association of specific data elements with additional descriptors (synonyms, external ontologies etc..) These process rules are driven by metadata generated during protocol design, and assay methodologies definition (SOP) and results in the annotation of both contextual and conceptual information to the raw data.
The repository contains specific data models design for each core assay (e.g. Flow cytometry). These multiple data models (databases) are unified through a set of common data tables that define the dimensions of patient demography, protocol (disease and therapy), and temporal concepts. We have intentionally not attempted to create a unified data model that reflects the biology of the assays; rather we look to establish a repository from which multiple data views can be created to feed biological model development.
Results: We will demonstrate the integration and visualization of these diverse data types through an on-line analytical processing OLAP tool. The system goal is the ability to browse the data within the repository with out the need for an understanding of the underlying data structure and then create adhoc data sets that can then be exported.