This article is part of the supplement: NIPS workshop on New Problems and Methods in Computational Biology

Open Access Highly Accessed Proceedings

ARACNE: An Algorithm for the Reconstruction of Gene Regulatory Networks in a Mammalian Cellular Context

Adam A Margolin12, Ilya Nemenman2, Katia Basso3, Chris Wiggins24, Gustavo Stolovitzky5, Riccardo Dalla Favera3 and Andrea Califano12*

Author Affiliations

1 Department of Biomedical Informatics, Columbia University, New York, NY 10032

2 Joint Centers for Systems Biology, Columbia University, New York, NY 10032

3 Institute for Cancer Genetics, Columbia University, New York, NY 10032

4 Department of Applied Physics and Applied Mathematics, Columbia University, New York, NY 10032

5 IBM T.J. Watson Research Center, Yorktown Heights, NY 10598

For all author emails, please log on.

BMC Bioinformatics 2006, 7(Suppl 1):S7  doi:10.1186/1471-2105-7-S1-S7

Published: 20 March 2006

Additional files

Additional File 1:

Determination of mutual information statistical significance. P-values are assigned to MI thresholds using a Monte Carlo simulation for different kernel widths, sample sizes (M) and for 105 gene pairs so that reliable estimates are produced up to p = 10-4 (solid lines). Extrapolation to smaller p-values is done using

<a onClick="popup('','MathML',630,470);return false;" target="_blank" href="">View MathML</a>

(dotted lines).

Format: EPS Size: 10KB Download file

Open Data

Additional File 2:

Prediction errors as a function of DPI tolerance. The number of inferred errors, NFP + NFN, are plotted as a function of the DPI tolerance, τ, for (a) the Erdös-Rényi and (b) the scale-free topologies. Raising τ to a value of 0.2 results in a modest increase in false positives, while larger values of τ produce a much sharper increase. Therefore, a moderate choice for the tolerance can help elucidate additional interactions without introducing an excessive number of false positives. Results are calculated for a statistical significance threshold of 10-4 and a synthetic microarray size of 1,000.

Format: PDF Size: 14KB Download file

This file can be viewed with: Adobe Acrobat Reader

Open Data

Additional File 3:

MI distribution for different shortest path lengths for the Erdös-Rényi topology. Red and black arrows are explained in the legend of Figure 5. Since there are no large in-degree hubs, decorrelation is slower than for the scale-free network, and MI statistics even for fifth neighbors is still distinguishable from the background.

Format: EPS Size: 61KB Download file

Open Data