Skip to contents

Transform a Table of Feature Abundances into a Table of Feature Set Abundances.

featureSetSummary.Rd

Represents a feature set by the mean or median feature measurement of a feature set for all features belonging to a feature set.

Usage

# S4 method for matrix
featureSetSummary(
  measurements,
  location = c("median", "mean"),
  featureSets,
  minimumOverlapPercent = 80,
  verbose = 3
)

# S4 method for DataFrame
featureSetSummary(
  measurements,
  location = c("median", "mean"),
  featureSets,
  minimumOverlapPercent = 80,
  verbose = 3
)

# S4 method for MultiAssayExperiment
featureSetSummary(
  measurements,
  target = NULL,
  location = c("median", "mean"),
  featureSets,
  minimumOverlapPercent = 80,
  verbose = 3
)

Arguments

measurements

Either a matrix, DataFrame or MultiAssayExperiment containing the training data. For a matrix, the rows are samples, and the columns are features. If of type DataFrame or MultiAssayExperiment, the data set is subset to only those features of type numeric.

location

Default: The median. The type of location to summarise a set of features belonging to a feature set by.

featureSets

An object of type FeatureSetCollection which defines the feature sets.

minimumOverlapPercent

The minimum percentage of overlapping features between the data set and a feature set defined in featureSets for that feature set to not be discarded from the anaylsis.

verbose

Default: 3. A number between 0 and 3 for the amount of progress messages to give. This function only prints progress messages if the value is 3.

target

If the input is a MultiAssayExperiment, this specifies which data set will be transformed. Can either be an integer index or a character string specifying the name of the table. Must have length 1.

Value

The same class of variable as the input variable measurements

is, with the individual features summarised to feature sets. The number of samples remains unchanged, so only one dimension of measurements is altered.

Details

This feature transformation method is unusual because the mean or median feature of a feature set for one sample may be different to another sample, whereas most other feature transformation methods do not result in different features being compared between samples during classification.

References

Network-based biomarkers enhance classical approaches to prognostic gene expression signatures, Rebecca L Barter, Sarah-Jane Schramm, Graham J Mann and Yee Hwa Yang, 2014, BMC Systems Biology, Volume 8 Supplement 4 Article S5, https://bmcsystbiol.biomedcentral.com/articles/10.1186/1752-0509-8-S4-S5.

Author

Dario Strbenac

Examples


  sets <- list(Adhesion = c("Gene 1", "Gene 2", "Gene 3"),
               `Cell Cycle` = c("Gene 8", "Gene 9", "Gene 10"))
  featureSets <- FeatureSetCollection(sets)
  
  # Adhesion genes have a median gene difference between classes.
  genesMatrix <- matrix(c(rnorm(5, 9, 0.3), rnorm(5, 7, 0.3), rnorm(5, 8, 0.3),
                        rnorm(5, 6, 0.3), rnorm(10, 7, 0.3), rnorm(70, 5, 0.1)),
                        nrow = 10)
  rownames(genesMatrix) <- paste("Patient", 1:10)
  colnames(genesMatrix) <- paste("Gene", 1:10)
  classes <- factor(rep(c("Poor", "Good"), each = 5)) # But not used for transformation.
  
  featureSetSummary(genesMatrix, featureSets = featureSets)
#> Summarising features to feature sets.
#>            Adhesion Cell Cycle
#> Patient 1  8.114219   4.958581
#> Patient 2  8.131557   5.006326
#> Patient 3  7.797197   4.861648
#> Patient 4  7.700979   5.082957
#> Patient 5  7.883901   4.987350
#> Patient 6  6.792416   5.112981
#> Patient 7  6.707971   4.987276
#> Patient 8  6.266468   4.950854
#> Patient 9  6.970349   5.027845
#> Patient 10 6.261045   4.960737