Statistics Seminar: Miss Kristen Feher -- Towards a method for resolving the dependence structure

Understanding gene expression networks is one of the key aims of
microarray exper- iments however many commonly used data analysis
techniques work on the assumption of independence among all
genes. Specifically, it can be difficult to get a complete picture using
correlation-based tools such as Expression Angler
(http://www.bar.utoronto.ca/), which uses a gene as input to return a
list of genes with correlation coeiffcients above a user-defined
threshold. Results from a gene used as input can in turn be used as
input and there is no rigourous procedure for choosing the
threshold. This means that this type of analysis can be subjective, and
a more objective treatment is needed. Random Matrix Theory (RMT) was
developed throughout the 1950s and 1960s to study and explain the
spectra of complex nuclei, and has since been used to explain other
complex systems as varied as disordered mesoscopic systems, financial
systems and traffic networks. In the context of microarray data, there
is promise that RMT will be able to unravel the genes that are dependent
on each other and the genes that are effectively independent of each
other under certain experimental conditions by examining the properties
of the eigenvalues of a gene-gene correlation matrix. This talk will
commence with a brief introduction to some important RMT results,
followed by a summary of possible output of RMT as applied to microarray
data.

Selected references:
Luo et al. (2006), Application of random matrix theory to microarray
data for discovering functional gene modules, Phys. Rev. E 73 

Luo et al. (2007), Constructing gene co-expression networks and
predicting functions of unknown genes by random matrix theory, BMC
Bioinformatics 8