nJSD

nJSD is a python package for calculating distance between two biological networks instantiated with gene-expression profiles using entropy concept. It was designed to measure intratumor heterogeneity from bulk RNA-sequencing data. Transcriptome-based ITH (tITH) of tumor state was calculated by considering both normal state and ideally heterogeneous state.

Installation

pip install njsd

Usage

nJSD supports command-line invocation as below:

usage: njsd [-h] -n NETWORK -r REF -q QUERY -o OUTPUT [-t GENESET]

Calculate network-based Jensen-Shannon Divergence.

optional arguments:
  -h, --help            show this help message and exit
  -n NETWORK, --network NETWORK
                        Pre-defined network
  -r REF, --ref REF     Reference gene expression profile
  -q QUERY, --query QUERY
                        Query gene expression profile
  -o OUTPUT, --output OUTPUT
                        Output file.
  -t GENESET, --geneset GENESET
                        Gene set list

Note that -t GENESET option is optional. If -t option is specified, gene set-specified nJSD and tITH will be computed. Otherwise, njsd will compute transcriptome-wide nJSD of the two expression profiles and tITH of query gene expression profile.

Network file, which should be given with -n/--network option must be formatted as below where each line specifies an edge in the network. njsd will simply ignore the header by skipping a single line, so you may name each column in a human-friendly way:

GeneA GeneB               # Header
GeneSymbol1 GeneSymbol2
GeneSymbol1 GeneSymbol3
GeneSymbol1 GeneSymbol4
...

Gene expression profile file must follow the format below. Again, the header doesn't matter. Note that njsd will automatically apply log2-transformation to expression values by taking log2(expression + 1), we recommend giving njsd unnormalized expression values, such as raw FPKM, RPKM or TPM.

GeneSymbol  ExpressionValue       # Header
GeneA 10
GeneB 20
BeneC 30
...

Gene set list file must follow the format below. Please be warned that this file should not have a header. The first column denotes names of each gene set(or group), and the following columns represent the member of each group.

Group1Name  GeneA   GeneB   GeneC   ...
Group2Name  GeneD   GeneE   GeneF   ...
Group3Name  GeneA   GeneG   GeneH   ...
...

When the gene set of reference GEP is differ to gene set of query GEP file and geneset list file. The difference is dumped into a file with name "dumpgene+date".

Toy Data

In the example directory, there are toy data.

example:

python nJSD.py whole -n example/Toy.network -r example/Toy.profile1 -i example/Toy.profile2

result:

example/Toy.profile2    [Ref. -> Query: 0.003935]       [Query -> stateH: 0.006820]     <tITH: 0.365869>

example:

python nJSD.py geneset -n example/Toy.network -r example/Toy.profile1 -i example/Toy.profile2 -t example/Toy.geneset

result:

example/Toy.profile2     1st_pwy         [Ref. -> Query: 0.007822]      [Query -> stateH: 0.009385]     <tITH: 0.454582> 
example/Toy.profile2     3rd_pwy         [Ref. -> Query: 0.005215]      [Query -> stateH: 0.007102]     <tITH: 0.423379> 
example/Toy.profile2     2nd_pwy         [Ref. -> Query: 0.000000]      [Query -> stateH: 0.004261]     <tITH: 0.000000> 
example/Toy.profile2     4th_pwy         [Ref. -> Query: 0.007909]      [Query -> stateH: 0.004261]     <tITH: 0.649850> 
example/Toy.profile2     5th_pwy         [Ref. -> Query: 0.004470]      [Query -> stateH: 0.012175]     <tITH: 0.268536> 

Citation

Y. Park, S. Lim, J. Nam, S. Kim, Measuring intratumor heterogeneity by network entropy using RNA-seq data, Scientific Reports (2016)