smooth-IWTomicsData: Smooth curves of a '"IWTomicsData"' object

In IWTomics: Interval-Wise Testing for Omics Data

Description

The function allows to smooth the curves in a "IWTomicsData" object, to fill in gaps in the measurements, to change measurement resolutions and, when region alignment is "scale", to measure the features in the different regions on the same grid.

Usage

## S4 method for signature 'IWTomicsData'
smooth(x, type="locpoly", 
  id_regions_subset=idRegions(x), id_features_subset=idFeatures(x),
  resolution_new=resolution(x)[id_features_subset], 
  scale_grid=unlist(lapply(lengthFeatures(x)[id_features_subset],
    function(length) max(unlist(length[id_regions_subset])))),
  fill_gaps=TRUE, bandwidth=5, degree=3, dist_knots=10, parallel=FALSE)

Arguments

x	"IWTomicsData" object.
type	type of smoothing used. Possible types are: "locpoly" for local polynomials (see help of locpoly function), "kernel" for Nadaraya-Watson kernel smoothing with Gaussian kernel (see help of ksmooth function), "splines" for regression b-splines (see create.bspline.basis and smooth.basis). Note that "splines" is very computational expensive when regions have different length and/or gaps.
id_regions_subset	vector with the identifiers of the region datasets to be smoothed.
id_features_subset	vector with the identifiers of the features to be smoothed.
resolution_new	resolution used to evaluate smoothed curves. As default, measurement resolution is mainteined after smoothing and smoothed curves are evaluated on the same grid as the raw curves. Can be either a constant (the same resolution for all features) or a vector of constants (a resolution for each feature). Not used if region alignment is "scale".
scale_grid	number of equally-spaced grid points over which the smoothed curves are evaluated. Can be either a constant (the same grid for all features) or a vector of constants (a grid for each feature). Only used if region alignment is "scale".
fill_gaps	if TRUE (default), smoothing is used to fill gaps (NA measurements). Can be set to FALSE only if region alignment is not "scale" and measurement resolution is mainteined.
bandwidth	bandwidth smoothing parameter used in "locpoly" or "kernel" type smoothing.
degree	degree of local polynomial or b-splines used in "locpoly" or "splines" type smoothing.
dist_knots	(approximate) distance between knots defining the b-splines used in "splines" type smoothing.
parallel	if TRUE smoothing is run in parallel on the different curves.

Value

smooth returns a "IWTomicsData" object with smoothed curves and updated resolutions and lengths (without the optional slot test).

Note

If the regions have different lengths and/or gaps are present in the feature measurements (NA measurements), "splines" type smoothing is very computational expensive. In this case we suggest to use "locpoly" or "kernel" type smoothing to smooth a large number of curves.

Author(s)

Marzia A Cremona, Alessia Pini, Francesca Chiaromonte, Simone Vantini

See Also

IWTomicsData for "IWTomicsData" class, constructors, accessors and methods; plot method to plot "IWTomicsData" objects; IWTomicsTest for the Interval-Wise Testing.

Examples

## -------------------------------------------------------------------------------------------
## -------------------------------------------------------------------------------------------
## CURVE ALIGNMENT CENTER
## -------------------------------------------------------------------------------------------
## -------------------------------------------------------------------------------------------
data(regionsFeatures_center)
dev.new()
plot(regionsFeatures_center,type='curves',N_regions=lengthRegions(regionsFeatures_center))

## Smooth all the curves with local polynomials
regionsFeatures_smooth=smooth(regionsFeatures_center,type='locpoly')
plot(regionsFeatures_smooth,type='curves',N_regions=lengthRegions(regionsFeatures_smooth))

## Smooth only feature 'ftr1', only region datasets 'elem3' and 'control'
regionsFeatures_smooth=smooth(regionsFeatures_center,type='locpoly',id_features_subset='ftr1',
                              id_regions_subset=c('elem3','control'))
plot(regionsFeatures_smooth,type='curves',N_regions=lengthRegions(regionsFeatures_smooth))

## Smooth only feature 'ftr1' and change its resolution
regionsFeatures_smooth=smooth(regionsFeatures_center,type='locpoly',id_features_subset='ftr1',
                              resolution_new=4000)
plot(regionsFeatures_smooth,type='curves',N_regions=lengthRegions(regionsFeatures_smooth))

## Smooth without filling gaps
regionsFeatures_smooth=smooth(regionsFeatures_center,type='locpoly',fill_gaps=FALSE)
plot(regionsFeatures_smooth,type='curves',N_regions=lengthRegions(regionsFeatures_smooth))


## -------------------------------------------------------------------------------------------
## -------------------------------------------------------------------------------------------
## CURVE ALIGNMENT SCALE
## -------------------------------------------------------------------------------------------
## -------------------------------------------------------------------------------------------
data(regionsFeatures_scale)
dev.new()
plot(regionsFeatures_scale,type='curves',N_regions=lengthRegions(regionsFeatures_scale),
                           average=FALSE,size=FALSE)

## Smooth all the curves with local polynomials
regionsFeatures_smooth=smooth(regionsFeatures_scale,type='locpoly')

## Plot of the smoothed curves
plot(regionsFeatures_smooth,type='curves',N_regions=lengthRegions(regionsFeatures_smooth),
     average=FALSE)

IWTomics documentation built on Nov. 8, 2020, 5:33 p.m.