smoothHeatmap: Smooth a heatmap

Description Usage Arguments Value Methods (by class) Examples

Description

Smooth a heatmap

Usage

1
2
3
4
5
smoothHeatmap(heatmap, ...)

## S4 method for signature 'Heatmap'
smoothHeatmap(heatmap, sigma = c(3, 3),
  output.size = dim(image(heatmap)), algorithm = NULL)

Arguments

heatmap

A heatmap object

...

additional arguments to S4 methods

This function smooths a heatmap using either binned kernel density (more efficient for binary heatmaps) or gaussian blur.

Sigma controls the SD of the kernel in both cases, defined in terms of pixels. This means that if you have very diffirent x and y dimensions (eg. a 200bp heatmap around 10000 promoters) you will need to compensate by setting sigma[2] higher to get the same visual effect in both dimensions

"output.size" specifies the dimensions of the output matrix. This can be useful to reduce plotting time significantly.

Smoothing can use either a kernel density estimate or a blurring function. The methods implemented are KernSmooth:bkde2D and EBImage::filter2 with a gaussian filter. The kernel based method assumes we are smoothing individual points so the value of these points are ignored. This is most useful for smoothing PatternHeatmaps where each cell in the matrix is either 1 or 0. For non-binary heatmaps, blur is most appropriate. Not setting this parameter will choose the method automatically.

Scaling the output heatmap is handled as in CoverageHeatmap.

sigma

Numeric, lengt2, (recycled if length 1)

output.size

Numeric, length 2

algorithm

"kernel" or "blur"

Value

A heatmap

Methods (by class)

Examples

1
2
data(HeatmapExamples)
hm_smoothed = smoothHeatmap(hm, sigma=c(5,5), algorithm="blur")

Example output

Attaching package: 'heatmaps'

The following object is masked from 'package:base':

    scale


Applying Gaussian blur...

heatmaps documentation built on Nov. 8, 2020, 7:32 p.m.