README file for s-core / s-core+ perl script

This README file describes how the s-core / s-core+ analysis perl script is to be executed together with data files.

Overview -------- This folder contains networks and code used for the s-core+ decomposition analyses presented in the manuscript "Comparative analysis of weighted gene co-expression networks in human and mouse", submitted to PLoS One, 2017. An example of how to use the perl script `s-core_plus.pl` is given below, under `s-core+`. In short: `perl s-core_plus.pl <input_network> <output_results_file>` It should be noted that the code is not optimized for speed or memory efficiency. It is provided to make the s-core / s-core+ method more transparent to the community. Weighted topological overlap (wTO) networks ------------------------------------------- There are 4 networks, 2 from mouse data, and 2 from human data. And one denoted by `all`, and one by `cns`, for both species. `all` means that the network is generated comparing gene-expression profiles across all tissue types in the data. `cns` means that only the central nervous system (CNS) and brain tissue types are included in the network. The `all` networks are a lot larger than the `cns` networks, since they contain >10,000 nodes, while the `cns` only contain transcription factors, and thus around 1,000 nodes. The networks are very dense undirected, weighted networks given as link lists. The naming convention is: `hs`: Homo Sapiens (Human) `mm`: Mus Musculus (Mouse) `all`: Network constructed from all tissue types `cns`: Network constructed from brain and CNS tissue types `cutX`: Cutoff value (`cut4`: cutoff = 0.4, `cut3`: cutoff = 0.3) s-core+ ------- The s-core method works by peeling off peripheral nodes in an undirected, weighted network. For unweighted networks, the s-core is equal to the k-core network decomposition method. s-core eventually peels the network down to an innermost core, and s-core+ commences. s-core+ works by removing the weakest links in the network until a new, stable innermost s-core can be found. In general, s-core is a relatively quick method, while s-core+ is slow. For instance, `s-core_plus.pl` decomposes the entire `all` networks into its s-cores in <30 minutes, while s-core+ takes a few days to complete. The `cns` networks are quick to run, both for s-core and s-core+, due to their smaller size and denseness. To, for instance, obtain the results for the human `cns` network, run: `perl s-core_plus.pl hs_wTO_cns_cut4.txt results.txt` The first argument of `s-core_plus.pl` is the <input_network>. The second argument of `s-core_plus.pl` is the <output_results_file>. The output is in the shape of 4 tab-delimited columns: `#Node HighestCore StrengthThreshold CoreSize` `Node`: Node ID (gene name) `HighestCore`: The innermost core `Node` is a member of `StrengthThreshold`: The strength threshold value for the given s-core(+) `CoreSize`: Number of nodes in core number `HighestCore` If only the s-core results are needed, and not the s-core+ results, adjust the parameter `$networkSizeForFallback` so that it is smaller than the innermost s-core (e.g. by setting it to 1). There is also an option for verbosity. Set `$ifVerbose = "yes"` if you want insight into the progress for every s-core iteration.