Cancer pathway Reconstruction using Breast Cancer Microarray gene expression profiles


Cell signaling pathways play a major role in the formation of cancer. In this task several gene expression profiles from 189 invasive breast carcinomas
were measured (Sotiriou et al. J. Natl. Cancer Inst. 2006). We here use the NetPath? ( cancer signaling pathway as evaluation criteria.

The dataset

Download (external link) (16 MB)

The dataset consistst of the 189 breast cancer expression profiles normalized using RMA and non-expressed genes were already filtered out.

X: expression data, each row corresponds to a gene, each column to a cancer expression profile
rownames: the gene symbols for each row of X
PathWay: struct of the pathway associated gene symbols

The task and evaluation criteria

The task is to reconstruct the signaling pathways from the microarray data using blind source separation techniques. As an underlying model we'd suppose a linear factor model as described in Lutter et al. BMC Bioinformatics, 2008, and Teschendorff et al. PLoS CB, 2007.

Evaluation of the reconstructed pathways will be done using Fisher´s exact test with multiple testing correction.


Each participant should sumbmit a list of ALL the reconstructed pathways. Each pathway should consist of a list of the involved genes and should be cleary labeled. Pathway lists can be submitted as a tab delimited csv file via email:
dominik.lutter (at)

Each participant is also asked to submit basic information about the used algorithm (dimension reduction, usage of prior knowledge, etc), and the total number of reconstructed pathways.

Potential participants

Florian Bloechl
Andreas Kowarsch

Task proposed by Biomedical/BCI Committee