The integration of the two Clinical Information Systems, the Genomic Information System, and the Gene Ontology (GO) database is achieved through the PrognoChip Mediator, which offers a virtual common query model while data is stored only in the constituent heterogeneous databases. The Mediator is developed using the ASP programming language and ODBC for accessing the corresponding databases.

The (authorized) biomedical investigator can form clinico-genomic queries through the web-based graphical user interface of the Mediator. This is split into several query forms, whose links are found in the left frame of the web interface. Through these forms, the user can specify criteria for selecting tumors:

• excised from patients with a clinical profile of interest
• having histopathologic characteristics of interest
• participating in a PrognoChip microarray experiment of specified quality and characteristics.

Returned tumors are accompanied with desirable clinico-genomic information and, in particular, their gene expression profiles. In all forms, numerical and datetime fields are queried as ranges. Additionally, some fields are dependent on a parent field and are appearing only if the parent field takes a specific value. After the selection criteria of a clinical or genomic form are saved, a complex SQL subquery, or part of it, is formed which is temporarily stored in the Mediator DB. The final SQL subqueries are formed only after the desired output fields are selected by the user. Then, the final subqueries are submitted to the corresponding Clinical ISs or Genomic IS for evaluation

Final results of the Clinical and the Genomic ISs are joined, based on the names of the cancerous samples, and an XML file, called output XML file, of predetermined schema is created. Since the set of selected reporters can be very large, reporter/gene annotations are stored in a separate tab-delimited file, called reporter file. Similarly, the requested normalized data fields of each cancerous sample are also saved (separately), in a tab-delimited file, called normalized data file. This way, the output XML file can contain only the names of the corresponding files, along with their column definition (eg., <reporter id, gene name (HUGO), Ensemble gene id, GO Biological Process> for the (single) reporter file, and <reporter id, log₂(intensity1/intensity2), log₂(intensity1 + intensity2)> for the normalized data files).

The output XML file is viewed through the current web browser, in a separate window, and the complete set of output files can be downloaded to the local machine of the user (as a .zip file). Then, it is given as input to the Data Mining tools for mining interesting clinico-genomic associations between the retrieved attributes of the selected samples. Finally, interesting Data Mining findings can be annotated and stored in the Findings Repository.

Presentation: ppt