---
title: "aracne.networks, a data package containing gene regulatory networks assembled from TCGA data by the ARACNe algorithm"
author:
- name: Federico M. Giorgi
  affiliation:
  - Department of Systems Biology, Columbia University, New York, USA
  - CRUK, Cambridge University, Cambridge, UK
- name: Mariano J. Alvarez
  affiliation:
  - Department of Systems Biology, Columbia University, New York, USA
  - DarwinHealth Inc., New York, USA
- name: Andrea Califano
  affiliation: Department of Systems Biology, Columbia University, New York, USA
date: "`r Sys.Date()`"
output: BiocStyle::html_document
vignette: >
  %\VignetteIndexEntry{Using aracne.networks}
  %\VignetteEngine{knitr::rmarkdown}
  %\VignetteEncoding{UTF-8}
---

```{r setup, include=FALSE}
knitr::opts_chunk$set(collapse = TRUE, comment = "#>")
```

# Overview of aracne.networks data package

The *aracne.networks* data package provides context-specific transcriptional
regulatory networks (also called interactomes or regulons) reverse engineered
by the ARACNe algorithm from The Cancer Genome Atlas (TCGA) RNAseq expression
profiles.

## ARACNe networks

This package contains `r nrow(aracne.networks::listRegulons())` Mutual
Information-based networks assembled by ARACNe-AP ([Giorgi et al., 2016](#Giorgi2016)) with default
parameters (MI p-value = $10^{-8}$, 100 bootstraps and permutation seed = 1).
ARACNe is a network inference algorithm based on an Adaptive Partitioning (AP)
Mutual Information (MI) approach ([Giorgi et al., 2016](#Giorgi2016)). In short, ARACNe-AP estimates
all pairwise Mutual Information scores between gene expression profiles, then
assesses the significance of such Mutual Information by comparison to a null
dataset. ARACNe then draws network edges between centroid genes (Transcription
Factors and Signaling Proteins) and genes significantly associated with them
(i.e. with significant MI). It then calculates Data Processing Inequality (DPI)
to reduce the number of indirect connections.

ARACNe-AP was run on RNA-Seq datasets normalized using Variance-Stabilizing
Transformation ([Anders and Huber, 2010](#Anders2010)). The raw data was downloaded on April 15^th^, 2015
from the TCGA official website ([Weinstein et al., 2013](#Weinstein2013)). We follow the TCGA naming
convention (e.g. BRCA = Breast Carcinoma) to name the individual
context-specific networks.

## Retrieving the networks

The networks are hosted on Zenodo
([doi:10.5281/zenodo.22918956](https://doi.org/10.5281/zenodo.22918956)).
The list of available networks is returned by `listRegulons()`:

```{r list}
library(aracne.networks)
listRegulons()
```

A network is retrieved with `getRegulon()`, using either its short name or the
name of the data set distributed with previous versions of the package (e.g.
`"regulonblca"`). The first call downloads the network and stores it in a local
cache managed by `r BiocStyle::Biocpkg("BiocFileCache")`; subsequent calls read
it from the cache, without requiring an internet connection.

```{r get}
regulonblca <- getRegulon("blca")
class(regulonblca)
length(regulonblca)
```

Code written for previous versions of the package, which used
`data(regulonblca)`, can be updated by replacing that call with
`regulonblca <- getRegulon("blca")`.

## Write a network to file

The package contains a function to print individual networks into a file.
Four columns will be printed: the Regulator id, the Target id, the Mode of
Action (MoA, inferred by Spearman correlation analysis ([Alvarez et al., 2016](#Alvarez2016))) that
indicates the sign of the association between regulator and target gene and
ranges between -1 and +1, the Likelihood (essentially an edge weight that
indicates how strong the mutual information for an edge is when compared to the
maximum observed MI in the network, it ranges between 0 and 1). Further details
about the *regulon* object as a model for transcriptional regulation are
present in the manuscript ([Alvarez et al., 2016](#Alvarez2016)).

In the following example, we print the first 10 interactions from the bladder
carcinoma (blca) network. The network genes are identified by Entrez Gene ids.

```{r write}
write.regulon(regulonblca, n = 10)
```

The user may want to analyze all the connections of a particular regulator
(E.g. "399", the RHOH gene).

```{r regulator}
write.regulon(regulonblca, regulator = "399")
```

# References

- <a id="Giorgi2016"></a>Giorgi, F.M. et al. (2016) ARACNe-AP: Gene Network
  Reverse Engineering through Adaptive Partitioning inference of Mutual
  Information. Bioinformatics 32(14):2233-2235. doi: 10.1093/bioinformatics/btw216.
- <a id="Anders2010"></a>Anders, S. and Huber, W. (2010) Differential
  expression analysis for sequence count data. Genome Biology 11(10):R106.
- <a id="Weinstein2013"></a>Weinstein, J.N. et al. (2013) The Cancer Genome
  Atlas Pan-Cancer analysis project. Nature Genetics 45:1113-1120.
- <a id="Alvarez2016"></a>Alvarez, M.J. et al. (2016) Functional
  characterization of somatic mutations in cancer using network-based inference
  of protein activity. Nature Genetics 48(8):838-847.

# Session information

```{r session}
sessionInfo()
```