DDHFm (version 1.1.4)

hftrialdatagen: Gene intensities simulator and DDHFm tester

Description

Simulates gene intensities and also applies DDHFm to them

Usage

hftrialdatagen(nreps = 4, nps = 128, plot.it = FALSE, uvp = 0.8)

Value

ansm

Two column matrix containing the estimated gene intensities and the true ones

hftssq

Sum of squares between estimated means and true means

yhf

Simulated gene intensities

Arguments

nreps

Number of replicates

nps

Number of genes

plot.it

Takes TRUE to activate the command of the respective plot and FALSE to deactivate it

uvp

a parameter for the denoising

Author

Guy Nason <[e-mail]>

Details

The code is well commented for further information.

First, genesimulator is called to obtain a vector of mean gene intensities (for a number of genes and a number of replicates for each gene.

Then link{simdurbin2} simulates a series of gene intensities using the (log-normal type) model as described in Durbin and Rocke (2001,2002).

Then for each gene the mean of replicates for that gene is computed.

Optionally, if plot.it is TRUE then the mean is plotted against its standard deviation (over replicates).

Then the intensities are sorted according to increasing replicate mean.

Optionally, if plot.it is TRUE then a plot of the intensities values as a vector (sorted according to increasing replicate mean) is plotted in black, and then the true mean plotted in colour 2 (on my screen this is red) and the computed replicate mean plotted in green.

The DDHF transform of the sorted intensities is computed.

Optionally, if plot.it is TRUE then a plot of the transformed means versus the transformed standard deviations is plotted. Followed by a time series plot of the transformed sorted intensities. These can be studied to see how well DDHF has done the transformation.

Then two smoothing methods are applied the the DDHF transformed data. One method is translation invariant, Haar wavelet universal thresholding. The other method is the classical smoothing spline. If plot.it is TRUE then these smoothed estimates are plotted in different colours.

Then the mean estimated intensity for each gene is computed and this is returned as the first column of a two-column matrix (ansm). The second column is the true underlying mean. The object hftssq contains a measure of error between the estimated and true gene means.

Examples

#
# First run hftrialdatagen
#
if (FALSE) v <- hftrialdatagen()
#
# Now plot the Haar-Fisz transformed intensities.
#
if (FALSE) ts.plot(v$yhf)
#
# Now plot the denoised intensities 
#
# Note that above we have 128 genes and 4 replicates and so there are
# 4*128 = 512 intensities to plot.
#
# However, there are only 128 gene intensities, and estimates. So, for this
# plot we choose to plot the noisy intensities and then for each replicate
# group (which are colocated on the plot) plot the (necessarily constant)
# true and estimated intensities (ie we plot each true/estimated intensity
# 4 times, once for each replicate).
#
# First estimates...
#
if (FALSE) lines(1:512, rep(v$ansm[,1], rep(4,128)), col=2)
#
# Now plot the truth
#
if (FALSE) lines(1:512, rep(v$ansm[,2], rep(4,128)), col=3)