Figure 1.
1) Data extracted from electronic health record (EHR) resources are semantically harmonized for data mining, generating a raw drug-event pair list. 2) The signal substantiation process analyses the submitted data, re-ranking the signal list, based on multiple algorithms. 3) Users trigger data analysis and exploration to validate the system operability.
Figure 2.
General architecture for the distributed pharmacovigilance platform.
Figure 3.
From user input to system output the platform engine controls the execution of workflows as follows:
1) One or more knowledge provider algorithms are selected to evaluate researcher-submitted datasets. The platform engine sends the request to the service execution engine. 2) An XML file with the input data (obeying the platform’s interoperability standard) is generated and its path provided to the service execution engine, along with the path for the workflows associated with each knowledge provider algorithm. The workflow execution is then triggered by a system call. 3) The Taverna command line tool loads the knowledge provider’s workflow, starting the processing tasks. 4) The knowledge provider execution proceeds internally, executing the miscellaneous workflow tasks. 5) The workflow delivers an XML data file (obeying the platform’s interoperability standard) with the algorithm output. 6) The service execution engine loads the XML output file and transfers the results to the platform engine. 7) The engine stores the data in the knowledge base and makes it promptly available for delivery in the web workspace.
Figure 4.
Internal platform implementation software overview.
Used components are implemented in Java, leveraging the open-source nature of this solution. 1) Platform engine components include Hibernate, JPA, Spring Security, POI, Log4j, Guice and custom code to control the application and serve it as a Tomcat web application. 2) The Web engine relies on Google Web Toolkit to generate a highly responsive web workspace. Add-ons such as GXT and Gin were used to improve the user interactions’ performance and reliability.
Figure 5.
EU-ADR Web Platform workspace interface for an undisclosed drug (XYZ) exploration scenario containing the signal list that results from distributed knowledge provider algorithm outputs and evidence combination statistical analysis.
Workflow results are labelled with Y in case sufficient evidence is found to support a potential drug-event relationship, or N otherwise. Evidence combination yields a score of H, M or L, indicating High, Moderate or Low risk respectively, of a drug-event relationship being in fact an ADR signal.