Gene interaction networks from donor and failing heart gene expression
Description
These are the gene-gene interaction networks obtained from gene expression microarray data in the paper "Pathologic gene network rewiring implicates PPP1R3A as a central cardioprotective factor in pressure overload heart failure" by Cordero, Parikh, et al. The dataset contains the following items:
* networks.zip: Compressed weighted matrices, in numpy format, for various methods in heart failure and donor cohorts, which are in the file name (ARACNE, JGL [Joint Graphical Lasso], Z-score, Pearson correlation, CLR [Context Likelihood of Relatedness])
* wgcna.zip: Compressed weighted matrices, in numpy format, for WGCNA. These are the networks used in the paper.
* network_gene_names.txt: Gene names for rows and columns of the gene networks in networks.zip
* wgcna
Construction of the networks were done as follows for each method:
* For Pearson, Z-score, and CLR, we simply computed each relevant statistic in a pairwise manner, no thresholds were applied.
* For ARACNE, we performed 500 bootstrapped iterations and searched the MI cutoff automatically using default parameters.
* For JGL we performed an Akaike Information Criteria guided model selection by scanning the sparsity and group difference L1 penalty strengths.