GSNA (version 0.1.4.2)

buildGeneSetNetworkLF: buildGeneSetNetworkLF, buildGeneSetNetworkLFFast-deprecated

Description

Using a gene set collection and a background of observable genes, calculate log partial Fisher p-value distances and return the results as a GSNData object. For a 2x2 contingency matrix of the form:

$$\biggl[\begin{matrix}a & b \\ c & d\end{matrix}\biggr]$$

the log Fisher p-value is equal to:

$$log(P) = log\biggl(\dfrac{(a+b)!(c+d)!(a+c)!(b+d)!}{a!b!c!d!(a+b+c+d)!}\biggr)$$

This differs from the buildGeneSetNetworkSTLF in that only the one value of P is summed, whereas in buildGeneSetNetworkSTLF, all more extreme values are summed (prior to log-transformation), generating an actual single-sided p-value.

This statistic behaves approximately like a 2-sided Fisher exact test, but may not be appropriate for most purposes. It is also somewhat faster to calculate than STLF (single tailed log-Fisher). Unless speed is an issue, we recommend using buildGeneSetNetworkSTLF Note: buildGeneSetNetworkLFFast is deprecated. Please use buildGeneSetNetworkLF() instead.

Usage

buildGeneSetNetworkLF(
  object = NULL,
  ref.background = NULL,
  geneSetCollection = NULL,
  distMatrixFun = function(geneSetCollection) scoreLFMatrix_C(geneSetCollection,
    alternative = 4)
)
buildGeneSetNetworkLFFast(...)

Value

This function returns a GSNData object with the $default_distance field set as 'lf' and $distances$lf$optimal_extreme set to 'min'.

Arguments

object

An object of type GSNData. If NULL, a new one is instantiated.

ref.background

(required) A character vector corresponding to the genes observable in a differential expression, ATAC-Seq or other dataset. This corresponds to the background used in tools like DAVID. This is required, unless object already exists and contains a genePresenceAbsence matrix field.

geneSetCollection

(required) A gene set collection either in the form of a tmod object, or a list of gene sets / modules as character vectors containing gene symbols and names corresponding to the gene module identifier. This is required, unless object already exists and contains a genePresenceAbsence matrix field.

distMatrixFun

Function used to calculate distances. Takes a genePresenceAbsence matrix and returns a distance matrix. (defaults to scoreLFMatrix_C)

...

: Arguments passed to buildGeneSetNetworkLF().

Functions

• buildGeneSetNetworkLFFast(): Deprecated, use buildGeneSetNetworkLF().

<code>buildGeneSetNetworkLFFast</code>

For buildGeneSetNetworkLFFast(), use buildGeneSetNetworkLF().

Details

This function wraps the process of creating a GSNData object and calculating a log Fisher p-value distance matrix. The distance matrix is calculated using scoreLFMatrix_C().

See Also

scoreLFMatrix_C scoreJaccardMatrix_C scoreOCMatrix_C

Examples

library(GSNA)
library(tmod)

# With tmod version >= 0.50.11, convert exported Bai_gsc.tmod **tmod** object to **tmodGS**:
if( utils::packageVersion( 'tmod' ) >= '0.50.11' )
  Bai_gsc.tmod <- tmod::tmod2tmodGS( GSNA::Bai_gsc.tmod )

# Get list of observable genes from expression data:
observable_genes <- toupper( rownames( Bai_empty_expr_mat ) )

# Subset GSEA data for significant results.
significant.Gsea <- subset( Bai_CiHep_dorothea_DN.Gsea, `FDR q-val` <= 0.05 )

# Subset tmod object for
gsc_subset.tmod <- Bai_gsc.tmod[ significant.Gsea$NAME ]

# Now, create a GSN object with partial log Fisher values:
GSN <- buildGeneSetNetworkLF( ref.background = observable_genes,
                              geneSetCollection = gsc_subset.tmod )