Tool description

PANDA reconstructs a gene regulatory network using TF PPI, TF DNA binding motif as regulation prior, and gene expression samples. PUMA reconstructs a gene regulatory network with miRNA as regulators using gene expression samples and miRNA predicted targets by miRanda or TargetScan as regulation priors. The Regulator-regulator interaction matrix is set to the identity matrix in the algorithm. LIONESS reconstructs patient-specific PANDA networks for each gene expression sample.

BONOBO is a single-sample network inference method, which can be used to estimate patient-specific co-expression matrices, which are then fed as input for PANDA to build patient-specific GRNs. To adapt BONOBO-PANDA networks for patient's sex, and in addition to sample-specific co-expression, PANDA's motif prior networks are constructed for each sex (accounting for sex chromosomes), while the third input (PPI networks) are generic for all patients. To download sample-specific networks, you can check the phenotypic information and select the networks by clinical variables or download all the samples in a single file.

Variable description

PANDA-LIONESS networks are single-sample networks generated by first estimating an aggregate PANDA network then deriving patient-specific LIONESS networks. You can either download all the networks to get a matrix of size the number of samples by the number of edges. The number of edges is 644 * 30243 (number of miRNA/TFs * number of genes) (~ 10⁶). Otherwise, you can specify the sample network to download and you will get a miRNA/TF-by-gene matrix named after the GTEx sample reference.