Biology Reference
In-Depth Information
For predictive purposes, it is useful to identify sets of cell lines that
share similar expression and growth phenotype patterns. An eventual
aim is to use such a compendium of sets in order to predict the growth
phenotype of a new sample based on the similarity of its expression
profile with a particular cell module (and suggest the adequate com-
bination of drugs to stop it from proliferating). In order to test our
predictions, we resorted to the public databases DrugBank and
Connectivity Map to show that the PPA predicts drug-gene associa-
tions significantly better than other methods. Moreover, comodules
not only have increased power to predict drug-gene associations,
but also significantly reduce the complexity of the data and provide
context in terms of the relevant experimental conditions. Our in-depth
analysis of a large compendium of comodules suggests that they
provide interesting new insights into possible mechanisms of
action for a wide range of drugs, and suggest new targets and novel
therapies.
3. Module Analysis
3.1. Module Annotation
The value of transcription modules or comodules depends critically on
what biological insight can be gleaned from the particular combina-
tion of elements (genes, samples, conditions, drugs, etc.) of which
they consist. With the growing body of information on the function
of gene products, 48,62,63 it is feasible to provide an initial module anno-
tation based on automated functional enrichment analysis. Specifically,
overrepresentation of genes belonging to the same functional category
in one module suggests its association with this function. Over-repre-
sentation can be quantified in terms of a p -value, based on the total
numbers of elements in the category and the module as well as
their intersection. Usually, these p -values are computed using Fisher's
exact test. Presently, several articles on online enrichment analysis
(FunSpec, 64 MAPPFinder, 65 and FatiGO 66 ) have been published.
Functional categories for many human and mouse genes are provided,
for example, by the Gene Ontology (GO) project, 67
and associations
Search WWH ::




Custom Search