Abstract
We apply a conditional independence-based network structure inference algorithm to the lung cancer dataset of Garber et al. Although network inference algorithms cannot help in reconstructing the complete causal network, we argue that they nevertheless can be used for extracting the network of i nfluences that are responsible for a particular expression profile anomaly, for which we have a reasonably large set of samples. This produces a sample-spec ific network that succinctly describes the dependencies among the variables in the given dataset. Run on the Garber dataset, our QFCI algorithm was able to reconstruct a plausible substructure of the ?small cell? subtype involving an expression profile typical for neuroendocrine differentiation. We propose usin g this type of approach as an alternative to clustering, which is very useful for tracking down groups of genes with similar expression profiles, but is una ble to infer the precise dependencies and (causal) mechanisms involving these genes.

URL:
http://rewerse.net/publications/rewerse-publications.html#REWERSE-RP-2004-88

BibTeX:

@inproceedings{REWERSE-RP-2004-88,
	author = {Liviu Badea},
	title = {Extracting networks of influences from microarray data (Poster)},
	booktitle = {Proceedings of 12th International Conference on Intelligent Systems for Molecular Biology, Glasgow, Scotland (31st July--4th August 2004)},
	year = {2004},
	organization = {International Society for Computational Biology},
	url = {http://rewerse.net/publications/rewerse-publications.html#REWERSE-RP-2004-88}
}