 
Summary: SUPPLEMENTARY MATERIAL
Comparing association network algorithms for reverse engineering
of large scale gene regulatory networks: synthetic vs real data
N. Soranzo, G. Bianconi and C. Altafini
SISSAISAS, International School for Advanced Studies
via Beirut 24, 34014 Trieste, Italy
Abdus Salam International Center for Theoretical Physics
Strada Costiera 11, 34014 Trieste, Italy
March 23, 2007
The material of this Supplement is divided into 3 Sections:
1. Synthetic data: integrates the content of Section 3.1 of the paper.
2. Influence of sparsity on the predictive power: compares inference on 2 networks with different
sparsity.
3. Comparing Bspline and Gaussian Kernel in the computation of I: evaluate how much the
matrix I changes with the algorithm chosen.
1 Synthetic data
This Section integrates the results obtained in Section 3.1 of the paper. For both AUC(ROC) and AUC(PvsR),
standard deviations (not shown) are around one order of magnitude smaller than the mean values, thus in
dicating that the repetitions are substantially faithful.
For the random and scalefree networks reconstructed in Fig. 1 of the paper, Fig. S1 reports the average
