Representing multidimensional gene expression data in databases

1
IBI SEMINAR SERIES
Representing multidimensional gene expression
data in databases Peter Vize Professor Departmen
ts of Biological Sciences, Computer Science
Biochemistry and Molecular Biology University of
Calgary Tuesday April 15, 2008 1200
Noon Biological Sciences 211 University of
Calgary Host Dr. Gordon Chua
Abstract Genes are expressed in complex patterns
over space and time. Computationally representing
and analyzing such data is difficult to achieve
but is an important undertaking if we are to be
able to predict protein-protein interactions and
local signal transduction networks in a
biologically meaningful way. I will describe two
very different methods of performing this type of
analysis in developing embryos. The first uses
computationally intensive methods to sample gene
expression data in 3 dimensions and to compare
and contrast different gene expression sets using
3D software analysis. In addition to the massive
data sampling associated with this approach,
embryo to embryo shape and size variation
interferes with the accuracy of this technique.
In order to avoid both of these issues we have
developed an alternative- a rough and dirty
solution using video gaming transformation tools.
A second approach uses no real computational
power but is very intensive in manpower. It
involves developing an anatomical ontology and
then using curators to apply the ontology to
manually evaluate gene expression data. Although
this last approach seems less powerful the
development of centralized reference ontologies
and related ontologies to describe phenotype and
mutant variation actually make this option far
more powerful for computational analysis of gene
function. We are now applying a highly developed
ontology to describe gene expression throughout
embryonic development based on a large scale
wholemount in situ analysis of over 3,000 genes.
Pizza and pop will be provided