Perform statistical tests on a cellHTS object

Share:

Description

This function takes a normalized, configured and annotated cellHTS object and performs statistical tests on it for the significance of a set of observations for each condition tested in a high-throughput screen.

Usage

1
2
3
cellHTS2OutputStatTests(cellHTSobject, annotationColumn = "GeneID",
controls = "neg", alternative = "two.sided", logged = FALSE, tests =
"T-test")

Arguments

cellHTSobject

an object of class cellHTS

annotationColumn

a single character value specifying the name of the column in the fData(cellHTSobject) data frame from which the feature identifiers will be extracted

controls

a single character value specifying the name of the controls to be used as a control population in the two-sample tests (this HAS to be corresponding to how these control wells have been annotated in the column "controlStatus" of the fData(cellHTSobject) data frame). If nothing is specified, the function will look for negative controls labelled "neg".

alternative

a single character value specifying the alternative hypothesis: "two.sided", "less" or "greater"

logged

a single logical value specifying whether or not the data has been logged during the normalization process

tests

a single character value specifying the tests to be performed: "T-test", "MannWhitney" or "RankProduct". If nothing is specified, all three tests will be performed. Be aware that the Rank Product test is slower than the other two, and returns a percent false discovery (equivalent to a FDR, not a p-value).

Details

The tests are computed taking into account only the wells labelled "sample" in the column "controlStatus" of the fData(cellHTSobject).

The two sample tests compare the set of observations for one construct to the values obtained for a population considered as "control". The one-sample tests compare the set of observations for one construct to the median of all values obtained across all constructs labelled as "sample". This type of test assumes that most constructs are expected to show a negligible effect. It is therefore not advised to use this type of tests when the constructs tested have been pre-screened for being associated with a phenotype.

Please be aware that both types of tests are less reliable when the number of replicates for each construct is low.

Value

a matrix with two columns, one for each type of test (two-sample and one-sample test) except the Rank Product (no alternative), and a row for each construct (row names corresponding to the identifiers given by the "annotationcolumn" entry).

Author(s)

Camille Terfve, Xin Wang

References

Michael Boutros, Ligia P. Bras L and Wolfgang Huber. Analysis of cell-based RNAi screens. Genome Biology 7:7 R66 (2006)."

Examples

1
2
3
4
5
6
7
8
9
## Not run: 
library(cellHTS2)
##load normalized cellHTS object (see the vignette for details about the
##preprocessing of this data set)
data("KcViab_Norm")
stats <- cellHTS2OutputStatTests(cellHTSobject=KcViab_Norm, alternative=
"two.sided", tests="T-test")

## End(Not run)

Want to suggest features or report bugs for rdrr.io? Use the GitHub issue tracker.