Significant interactions, SNP-Gene Interactions, Node and Edge Weights, cluster assignments, transitioning gene sets and network figures are available at the website: https://pages.discovery.wisc.edu/~bbaur/Roadmap_RegulatoryVar/
note: Additional readmes and examples are provided in each of these directories.
#Feature Generation: Code, examples and information on feature generation is provided in the genFeaturesForDiscrete data directory here.
#Prediction Generation: The random forest approach to train and predict contact counts can be found in the LHiCReg folder. Please see the README in that folder for usage and tips! L-HiC-Reg simply uses features from 1 Mb regions as input into HiC-Reg. We used the code in generateLocalPredictions (see README there) to extract the features in a 1 Mb region from the whole of the output from the feature generation code.
#Significant interaction calling: The code for calling significant interactions with the binomial method is provided in the sigcallinter directory.
#Scoring nodes based on significant interactions with SNPs: The code for mapping significant interactions to SNPs and scoring the genes based on their significant interactions with SNPs is provided in the mapregionpair_snp directory. This is what is used as input into the graph diffusion and MTGC pipeline.
#Network Generation: The code for generating the cell-type specific networks is included in the NetworkGeneration directory. Note that this includes the distal and proximal networks.
#Graph diffusion and eigenvector calculation: Information and code on the diffusion process and eigenvector calculation is provided in DiffusionAndEigenvectors
#Muscari After the eigegnvectors are calculated, to do the multi-task clustering we used MUSCARI https://github.com/Roy-lab/Muscari
#Transitioning gene sets: Transitioning gene sets were produced with the de novo clustering approach in: https://github.com/Roy-lab/clade-specific_gene_sets