GRNdata
In grndata: Synthetic Expression Data for Gene Regulatory Network Inference

This package contains a large set of gene expressions generated by various simulators collected in what we cal ``Datasource".

The data generated by the simulators is free of noise. The noise could be added later so that it is possible to control its properties independently of the simulators and also to provide fully reproducible tests. This study involves data generated by three different GRN simulators:

GNW

The GNW simulator [@schaffter2011genenetweaver] generates network structures by extracting parts of known real GRN structures capturing several of their important structural properties. To produce gene expression data, the simulator relies on a system of non-linear ordinary differential equations (ODE).

SynTReN

The SynTReN simulator [@van2006syntren] generates the underlying networks by selecting sub-networks from \emph{E.coli} and \emph{Yeast} organisms. Then the experiments are obtained by simulating equations based on Michaelis-Menten and Hill kinetics under different conditions.

Rogers

The data generator described in [@Rogers15072005] that will be referred as \emph{Rogers} relies on a power-law distribution on the number of connections of the genes to generate the underling network. The steady state of the system is obtained by integrating a system of differential equations simulating only knockout data.

Datasources

Using these simulators, five large datasources involving many noise-free experiments have been generated. The characteristics of these datasources are detailed in the following Table:

| Datasource | Topology | Experiments | Genes | Edges | |:----------------:|:-----------------------:|:-----------:|:-----:|:-----:| | $Rogers_{1000}$ | Power-law tail topology | 1000 | 1000 | 1350 | | $SynTReN_{300}$ | E. coli | 800 | 300 | 468 | | $SynTReN_{1000}$ | E. coli | 1000 | 1000 | 4695 | | $GNW_{1565}$ | E. coli | 1565 | 1565 | 7264 | | $GNW_{2000}$ | Yeast | 2000 | 2000 | 10392 |

In order to generate these datasources we have simulated multifactorial data with SynTReN and GNW, which is a less informative data [@marbach2010revealing].