Biology Reference
In-Depth Information
the variable hierarchy described above. Care must be taken to avoid
unspecified dependencies between samples.
4.2. Aim of SwissBrod
To avoid these problems and facilitate data mining and integration while
ensuring high data quality, the authors work with SwissBrod, an in-house
database developed at the Swiss Institute of Bioinformatics (SIB). The
aim of SwissBrod is to provide curated clinical and expression data in a
form ready for detailed statistical meta-analysis. 4,7 Aiding correct applica-
tion and interpretation of statistical methods is the main goal. This
involves identifying the actual sampling units (patients, tissues, or arrays)
and design (patient selection criteria).
In line with the precise scope of our aims, SwissBrod contains pri-
mary data on breast cancer only, and is currently limited to patient-based
tumor data. Data normalization was provided by the original study
authors. No other derived or summary results or statistical analyses are
provided, as these may be computed at will by users.
4.3. SwissBrod Data Curation
Primary datasets are acquired from a variety of public sources: tables
within journal articles, supplementary materials on a journal website,
author websites, and public repositories (e.g. GEO, ArrayExpress,
Stanford Microarray Database). As part of the curation process, the
datasets undergo quality controls and some reconfiguration to facilitate
downstream analyses.
A major aim of the curation process is to redefine and consolidate
dataset variables and cohorts in a way appropriate for statistical analysis.
The new cohorts must satisfy basic conditions, such as independent
patients as sampling units, and may cut across articles and original dataset
boundaries, as happens when a study is extended by the same investigat-
ing group. Study design is noted, along with selection criteria. In addition,
stable probe identifiers are established so that mapping of probes to
genes can be readily carried out.
Search WWH ::




Custom Search