Biomedical Engineering Reference
In-Depth Information
through two paradigms: a search interface and a hypothesis-testing interface.
The search engine allows users to create queries through combinations of ele-
ments from a biological subject dictionary, including pathways, genes, diseases,
trials, and compounds. The analysis engine (called Dataset Explorer) provided
utilities for analyzing phenotypes and genomic data at a cohort level, including
trial observations and endpoints. In the following sections we describe in
greater depth the functionality of these two engines.
16.7.1
Dataset Explorer
Dataset Explorer provides the scientists and physicians with a unique in silico
hypothesis-testing facility. The i2b2 software [17] was used as the basis for
this component, but key modifi cations and additions were made. The users
can create virtual cohorts using characteristics from a predefi ned proprietary
ontology. Not all data fi elds were comparable across studies, so the onto-
logy was organized by study. Within each study, information was categorized
using a common structure, including demographic information, clinical data,
sample data, and biomarker information (Fig. 16.4). For unique data elements
which are comparable across studies we developed a separate ontology
and user paradigm to enable cross-comparing multiple studies for a given
phenotype.
We created a selection tool for defi ning multiple cohorts and rapidly estab-
lishing statistical comparisons between those cohorts using a
t
- test or
2
-
statistics and visualizing the results using pie charts, histograms, and box plots
(Figs. 16.5 and 16.6). Additionally, we implemented components for querying,
viewing, and comparing the associated biomarker information such as gene
expression, proteomics, rules-based medicine protein panels, metabolomics,
and SNP data. Simple (Fig. 16.7) and clustered heatmaps are provided for
viewing expression data through the GenePattern tool [18], and linkage dis-
equilibrium maps for viewing SNP data are implemented through the
Haploview application [21] .
A comparative marker selection [22] module of GenePattern is deployed
to develop biomarker hypotheses, and a meta-analysis across multiple gene
expression experiments is also enabled for data sets which were measured on
comparable platforms (Fig. 16.8 ).
The system is designed to be able to meet the needs of the casual user so
it is envisioned that there are cases where the analysis capabilities imple-
mented are not suffi cient for the scientifi c task at hand. Therefore the aligned
and cleaned data can easily be exported for more in-depth analysis.
Finally, we added fi ne-grained access control to fi t our environment as
described above. Thus each user has a different view of the study tree accord-
ing to their access rights: the ones they are granted access by the study owners
can be opened in the tree view for exploration but the ones that have no access
rights can only see the metadata. Everyone has access rights to the public
studies.
χ
Search WWH ::
Custom Search