Skip to main content

Architecture

A key enabler of HIV pharmacogenomics research in Europe is a sound biomedical information integration platform to integrate and correlate the data. This data is described in diverse formats and vocabularies and scattered in disparate countries throughout Europe.

We utilize a platform that is based on three conceptual layers: integration, index and prediction models.

Conceptual Architecture Layers
Conceptual Architecture Layers

The Integration Layer transforms ARCA, AREVIR and Karolinska data into standard-based formats and integrates the normalized data. The integration layer employs data processing modules such as transformations from propriety formats to standard XML-based representation, alignment of the various vocabularies, data cleansing, etc. One key research area in this layer is to find the best standard to represent the data and enhance it to support the EuResist requirements. Those enhancements will be then fed back to the standard organizations, so they can improve their standards specifications. Some potential standards are HL7 Clinical Document Architecture (CDA), HL7 Clinical-Genomic message, and Bioinformatic Sequence Markup Language (BSML).

The Index Layer then provides query and search capabilities to the integrated data. There are two options for integrating the data. One is the aggregation model in which the data is stored in a repository such as a relational database and the indexation and queries are performed on that repository. The other option is the federation model in which no data is stored outside from the sources. Instead, at query time, the query is divided into sub-queries, which are posed against the various data sources. The corresponding sub-results are then jointed to one result to form the query result. In the federation model there is no data stored outside of the sources premises, but on the other hand it requires that the sources are on-line all the time and the performance of the queries is decreased with respect to the aggregation model. A key research area in this layer is to design a complete and efficient data model to support the prediction algorithms.

The Prediction Models Layer includes all the algorithms that can predict HIV medication efficacy in the presence of new HIV sequence variations.

Service Oriented Architecture
We are utilizing a Service Oriented Architecture (SOA) to implement the platform. SOA enables flexible integration of applications or resources such as databases by representing every application or resource as a service with a standardized interface connecting to a service integration bus. This allows the quick combination of new and existing databases and a staged development of the full system. Figure 2 below depicts the various services attached to the EuResist bus.