NETPAGE (NETwork Propagation-based Assessment of Genetic Events) is a novel computational framework for gene-based association testing of rare variants that integrates prior knowledge about tissue-specific gene interaction networks (Scelsi et al., PLOS Comp Biol 2021, in press; DOI: 10.1371/journal.pcbi.1008517).
The framework combines network propagation with sparse regularised regression. Here we provide a Python module to perform network propagation.
These instructions will get you a copy of the project up and running on your local machine for development and testing purposes.
You will need Python 2.7 to use this module. A list of module dependencies can be found in the requirements.txt
file.
No installation is required. Just clone this repo to your machine, make sure you have all the packages needed, and you're ready to go!
python RunNetworkSmoothingParser.py -h
usage: RunNetworkSmoothingParser.py [-h] [-v] -i INPUTNET -g GENEBURDEN --typ
{binary,burden} --scope
{entrezgene,ensembl.gene,ensembl.protein}
[-o OUTCOME]
[-e {void_protectives,void_risks,invert_protectives,logOR,none}]
[-t TOL] [-a ALPHA] [-s] [--binarynet]
[-p PERCENTAGE] [--edgeremoval]
[--probremoval PROBREMOVAL] [-r]
[--nrand NRAND] [--quantile]
[--pathway PATHWAY] [--plots]
optional arguments:
-h, --help show this help message and exit
-v, --version show program's version number and exit
-i INPUTNET, --inputnet INPUTNET
Path and name of the input network file (required)
-g GENEBURDEN, --geneburden GENEBURDEN
Path and name of the gene burden table file (required)
--typ {binary,burden}
Type of rare variant encoding (options allowed: binary, burden; required)
--scope {entrezgene,ensembl.gene,ensembl.protein}
Gene naming convention used in the network file.
Needed to map gene IDs to HGNC symbols.
-o OUTCOME, --outcome OUTCOME
File storing the outcome wrt which direction of effect must be considered (required)
-e {void_protectives,void_risks,invert_protectives,logOR,none}, --effect {void_protectives,void_risks,invert_protectives,logOR,none}
How do you want to deal with direction of effects? (case-sensitive, default "none")
* "void_protectives": set to 0 the mutation status/burden of protective genes
* "void_risks": set to 0 the mutation status/burden of risk genes
* "inversion": set mutation status to -1 for protective, +1 for risk genes
* "logOR": replace mutation status with logOR from Fisher test
* "none": do nothing
-t TOL, --tol TOL Convergence threshold for network smoothing (float, default 1e-6)
-a ALPHA, --alpha ALPHA
Diffusion distance allowed for a mutation signal through the network (float, default 0.5).
Multiple values not allowed
-s, --selfloops Should the network have self loops? (default False)
--binarynet Binarize the weighted network to an adjacency matrix;
pick only the top p% edges, as supplied with -p (default False)
-p PERCENTAGE, --percentage PERCENTAGE
Percentage of top edges to retain when binarising the network (float, default 1)
--edgeremoval For each gene, remove a random edge with probability p-rem,as supplied with --prob_removal (default False)
--probremoval PROBREMOVAL
Probability of randomly removing an edge incident on a gene (default 0.1)
-r, --randomnet If present, perform degree-preserving randomisation of the network after binarisation
--nrand NRAND, -n NRAND
How many replicates of the randomised network? (default 30)
--quantile, -q Quantile-normalise the smooth profile's rows (default False)
--pathway PATHWAY, -pw PATHWAY
Restrict network and mutation burden to genes in a pathway.
Please provide name of the file containing the list of genes in the desired pathway
(one gene per line)
--plots, -pl Plot a heatmap of the network? (default False)