Downstream analysis for MERLIN inferred networks
randpartitions_with_transpose is used to create sub-samples.
You can run it on an expression matrix with:
- genes on rows and samples on columns (normal) or
- samples on rows and genes on columns (transpose). It assumes there is no header file. So:
- For normal, first column is gene names, there is no sample name in first line
- For transpose, first row is gene names, there is no sample name in first column For example:
./makePartitions trans.txt 100 outdir/ 50 rand transpose
- trans.txt is a transposed expression matrix
- it creates 100 subsamples (dataset0.txt to dataset99.txt)
- in outdir/
- each with 50 randomly selected samples.
estimateedgeconf creates a consensus network from a list of network files:
./estimateEdgeConf network_files.txt 0 output_net_ alledges
where network_files.txt has the location of individual network files and the output (output_net_alledge.txt) will contain edges, and percentage of times the edges were seen in individual networks (1 means 100%).
assessclusterstab creates a co-clustering matrix that shows how many times two genes were in the same module.
./assessClusterStab module_files.txt sims.txt
where module_files.txt has the list of module assignments files (one per line) and sims.txt will be the co-clustering matrix.
optimalleaforder applies hierarchical clustering to co-clustering matrix from the previous step.
./reorder sims.txt matrix consensus_module_0.3 0.3
where 0.3 is the threshold for creating the hierarchical modules.