Seminar Details
2007-05-08 (14.00) : Identifying novel therapies for breast cancer using Independent Component Analysis
At Auditoire Euler
Organized by Mathematical Engineering
Speaker :
Andrew Teschendorff
(Cambridge University Hospitals
)
Abstract :
Most standard methods for analysing tumour-derived gene expression
data do not attempt to explicitly infer the altered biological processes underlying
cancer. By viewing gene expression as a blind source separation (BSS) problem we
can characterise the inferred biological processes in terms of aberrations in functional
pathways and transcriptional programs. Using Independent Component Analysis (ICA)
to perform BSS , we show that ICA significantly outperforms other linear decomposition techniques.
We describe the application of ICA in a meta-analysis of breast cancer, leading to novel associations
between biological pathways, regulatory modules and breast cancer phenotypes.