DaMiR.SV: Identification of Surrogate Variables

In BioinfoMonzino/DaMiRseq: Data Mining for RNA-seq data: normalization, feature selection and classification

Description

This function returns a matrix of surrogate variables (sv) using the implementation by Chiesa-Piacentini or the sva method by Leek et al.

Usage

DaMiR.SV(
  data,
  method = c("fve", "leek", "be"),
  th.fve = 0.95,
  second.var = NULL
)

Arguments

data	A SummarizedExpression object
method	The method used to identify sv. If missing, the "fve" method will be selected. Otherwise the method "leek" or "be" should be choosen
th.fve	This argument sets the threshold of maximum fraction of variance explained (fve) to be used in conjunction with "fve" method; default is 0.95
second.var	A factor or a numeric vector corresponding to an additional variable to take into account during the sv identification. This variable together with 'class' in the data object will be used to design the model matrix (~ class + second.var)

Details

This function helps the user to identify the appropriate number of sv: it is possible to select a different strategy to be used by changing the option in method argument. Three methods are available:

• "be" - this option uses the num.sv function of sva package with default parameters;

• "leek" - The same of before but with asymptotic approach proposed by Leek;

• "fve" - This method is introduced in DaMiRseq package, and integrates part of sva function with custom code. Briefly, we computed eigenvalues of data using code already implemented in sva function and then, we calculated the squared of each eigenvalues. Thus, the ratio between each "squared eigenvalue" and the sum of them were calculated. These values represent a surrogate measure of the "Percentage of Explained Variance" (pve) obtained by principal component analysis (PCA), and their cumulative sum can be used to select sv.

Value

A matrix of sv. A plot with the sv identified by "fve" method is also returned. A red dot shows the maximum number of variables to be included for a specific "fve".

Author(s)

Mattia Chiesa, Luca Piacentini

References

Jeffrey T. Leek, W. Evan Johnson, Hilary S. Parker, Elana J. Fertig, Andrew E. Jaffe and John D. Storey (2016). sva: Surrogate Variable Analysis. R package version 3.22.0.

See Also

sva

Examples

# use example data:
data(data_norm)
sv <- DaMiR.SV(data_norm, method = "fve", th.fve=0.95)

BioinfoMonzino/DaMiRseq documentation built on Aug. 22, 2021, 3:11 p.m.