This GitHub repository contains the code to reproduce the networks modeled in the LIONESS STAR Protocols manuscript. The code and files are structured into two main folders. The first folder is titled lioness_networks and includes files and scripts to run the LIONESS networks. The second folder is titled visualization and includes scripts and data to reproduce Figure 2.
The code that is included in lioness_networks models four types of networks—networks modeled on data from human and yeast, using either Pearson correlation or PANDA as aggregate network reconstruction algorithms. We separated the files into two folders, one titled human and one titled yeast. For both organisms, we include two folders with data and scripts, one for the LIONESS networks modeled on Pearson correlation (folder pearson) and one for the LIONESS networks modeled on PANDA (folder panda).
The Pearson correlation networks are modeled with lionessR. The data that is included (file exp.txt) contains the 1000 most variable genes. The networks are be modeled with a simple R script titled script.R. The output is saved into a file titled lionessResults.RData.
The PANDA networks are modeled on the gene expression data that we presented in the LIONESS paper, as well as a prior network of putative transcription factor-DNA binding (a "motif" prior) and protein-protein interactions ("ppi") between transcription factors. These files can be found in the input folder. This folder also includes a sampleorder.txt file, that is used to combine sample names with the output networks. In addition to the input folder, the panda folder includes a lib folder with functions that are needed to run PANDA, a mat folder where the aggregate PANDA networks will be saved, an output folder where the LIONESS networks will be saved in .mat format, and a txt folder where the networks will be saved in .txt format.
The networks can be modeled by executing the run_read_combine.sh script. This script executes multiple scripts. It starts modeling an aggregate panda network by executing panda_run.sh, then runs each of the individual sample networks with lioness_run.sh. After this, it reads the MATLAB output files and prints them as .txt files (read_and_print_networks.sh), which are then loaded into R and combined into one matrix with edge information in combine_networks.R. The output is saved into a file titled lionessResults.RData.
The folder visualization contains code to reproduce Figure 2.
(1) ExtractData.R: Convert the network files from RData to txt.
(2) VizInitialData4.m: Makes figures visualizing Pearson networks.
(3) VizInitialData_PANDA4.m: Makes figures visualizing the PANDA networks.
(4) SmoothScatter.R: Make smooth scatter plots for Pearson and PANDA networks in R.
(1) Data: contains the original yeast gene expression data and the aggregate yeast PANDA network; where extracted networks will be written.
(2) Tools: contains additional matlab packages for the visualization.
(3) Figure: where the figures will be saved.