ADaCGH2 (version 2.12.0)

pChromPlot: Segment plots for aCGH as PNG

Description

Produce PNG figures of segment plots (by chromosome) for aCGH segmentation results. Internal calls are parallelized for increased speed and we use ff objets to allow the handling of very large objects. The output can include files for creating HTML with imagemaps.

Usage

pChromPlot(outRDataName, cghRDataName, chromRDataName, probenamesRDataName = NULL, posRDataName = NULL, imgheight = 500, pixels.point = 3, pch = 20, colors = c("orange", "red", "green", "blue", "black"), imagemap = FALSE, typeParall = "fork", mc.cores = detectCores(), typedev = "default", certain_noNA = FALSE, loadBalance = TRUE, ...)

Arguments

outRDataName

The RAM object or the RData file name that contains the results from the segmentation (as an ffdf object), as carried out by any of the pSegment functions.

Note that the type of object in outRDataName, cghRDataName, chromRDataName, posRDataName, should all be of the same type: all ff objects, or all RAM objects.

As well, note that if you use RAM objects, you must use typeParall = "fork"; with ff objects you can use both typeParall = "cluster" and typeParall = "fork". Further details are provided in the vignette.

cghRDataName

The Rdata file name that contains the ffdf with the aCGH data or the name of the RAM object with the data.

If this is an ffdf object, it can be created using as.ffdf with a data frame with genes (probes) in rows and subjects or arrays in columns. You can also use inputToADaCGH to produce these type of files.

chromRDataName

The RData file name with the ff (short integer) vector with the chromosome indicator, or the name of the RAM object with the data. Function inputToADaCGH produces these type of files.

posRDataName

The RData file name with the ff double vector with the location (e.g., position in kbases) of each probe in the chromosome, or the name of the RAM object with the data. Function inputToADaCGH produces these type of files.

This argument is used for the spacing in the plots. If NULL, the x-axis goes from 1:number of probes in that chromosome.

probenamesRDataName

The RData file name with the vector with the probe names or the RAM object. Function inputToADaCGH produces these type of files. (Note even if this is an RData file stored on disk, this is not an ff file.) This won't be needed unless you set imagemap = TRUE.

imgheight

Height of png image. See png.

pixels.point

Approximate number of pixels that each point takes; this determines also final figure size. With many probes per chromosome, you will want to make this a small value.

pch

The type of plotting symbol. See par.

colors

A five-element character vector with the colors for: probes without change, probes that have a "gained" status, probes that have a "lost" status, the line that connects (smoothed values of) probes, the horizontal line at the 0 level.

imagemap

If FALSE only the png figure is produced. If TRUE, for each array * chromosome, two additional files are produced: "pngCoord_ChrNN@MM" and "geneNames_ChrNN@MM", where "NN" is the chromosome number and "MM" is the array name. The first file contains the coordinates of the png and radius and the second the gene or probe names, so that you can easily produce an HTML imagemap. (Former versions of ADaCGH did this automatically with Python. In this version we include the Python files under "imagemap-example".)

typeParall

One of "fork" or "cluster". "fork" is unavailable in Windows, and will lead to sequential execution. "cluster" requires having set up a cluster before, with appropriate calls to makeCluster, in which case the cluster can be one of the available types (e.g., sockets, MPI, etc).

Using "fork" and "cluster" will lead to different schemes for parallelization. See the vignette.

If you use ff objects, you can use different options for typeParall for segmentation and plotting.

mc.cores

The number of cores used if typeParall = "fork". See details in mclapply

typedev

The device type. One of "cairo", "cairo-png", "Cairo", or "default". "Cairo" requires the Cairo package to be available, but might work with headless Linux server without png support, and might be a better choice with Mac OS. "default" chooses "Cairo" for Mac, and "cairo" otherwise.

certain_noNA

Are you certain, absolutely sure, your data contain no missing values? (Default is FALSE). If you are, you can achieve considerable speed ups by setting it to TRUE. See the help for this option in pSegment. Of course, if you are setting it to true, the object you pass with the output, outRDataName, must have been generated using certain_noNA.

loadBalance

If TRUE (the default) use load balancing with MPI (use clusterApplyLB instead of clusterApply) and a similar approach for forking ( set mc.preschedule = FALSE in the call to mclapply ).

...

Additional arguments; not used.

Value

Used only for its side effects of producing PNG plots, stored in the current working directory (getwd().)

See Also

pSegment

Examples

#####################################################
###
### Using forking with RAM objects
###
#####################################################

### Note to windows users: under Windows, this will
### result in sequential execution, as forking is not
### available.


## Get example input data and create data objects

data(inputEx)

## (this is not necessary, but is convenient;
##  you could do the subsetting in the call themselves)
cgh.dat <- inputEx[, -c(1, 2, 3)]
chrom.dat <- as.integer(inputEx[, 2])
pos.dat <- inputEx[, 3]



## Segment with HaarSeg
haar.RAM.fork <- pSegmentHaarSeg(cgh.dat, chrom.dat,
                                   merging = "MAD")


pChromPlot(haar.RAM.fork,
           cghRDataName = cgh.dat,
           chromRDataName = chrom.dat,
           posRDataName = pos.dat,
           imgheight = 350)


## Not run: 
# 
# #####################################################
# ###
# ### Using a cluster with ff objects and create imagemaps
# ###
# #####################################################
# 
# 
# 
# ## Create a temp dir for storing output
# dir.create("ADaCGH2_plot_tmp_dir")
# originalDir <- getwd()
# setwd("ADaCGH2_plot_tmp_dir")
# 
# 
# ## Start a socket cluster. Change the appropriate number of CPUs
# ## for your hardware and use other types of clusters (e.g., MPI)
# ## if you want.
# 
# cl2 <- makeCluster(4,"PSOCK")
# clusterSetRNGStream(cl2)
# setDefaultCluster(cl2) 
# clusterEvalQ(NULL, library("ADaCGH2"))
# ## The following is not really needed if you create the cluster AFTER
# ## changing directories. But better to be explicit.
# wdir <- getwd()
# clusterExport(NULL, "wdir")
# clusterEvalQ(NULL, setwd(wdir))
# 
# 
# ## Get input data in ff format
# ## (we loaded the RData above, but we need to find the full path
# ## to use it in the call to inputToADaCGH)
# 
# fname <- list.files(path = system.file("data", package = "ADaCGH2"),
#                      full.names = TRUE, pattern = "inputEx.RData")
# 
# inputToADaCGH(ff.or.RAM = "ff",
#               RDatafilename = fname)
# 
# 
# 
# ## Segment with HaarSeg
# 
# haar.ff.cluster <- pSegmentHaarSeg("cghData.RData",
#                                    "chromData.RData",
#                                    merging = "MAD",
#                                    typeParall= "cluster")
# 
# ## Save the output (an ff object) and plot
# save(haar.ff.cluster, file = "haar.ff.cluster.out.RData",
#      compress = FALSE)
# 
# 
# pChromPlot(outRDataName = "haar.ff.cluster.out.RData",
#            cghRDataName = "cghData.RData",
#            chromRDataName = "chromData.RData",
#            posRDataName = "posData.RData",
#            probenamesRDataName = "probeNames.RData",
#            imgheight = 350,
#            imagemap = TRUE,
#            typeParall= "cluster")
# 
# ### Explicitly stop cluster
# stopCluster(NULL)
# 
# ### Clean up (DO NOT do this with objects you want to keep!!!)
# load("chromData.RData")
# load("posData.RData")
# load("cghData.RData")
# 
# delete(cghData); rm(cghData)
# delete(posData); rm(posData)
# delete(chromData); rm(chromData)
# unlink("chromData.RData")
# unlink("posData.RData")
# unlink("cghData.RData")
# unlink("probeNames.RData")
# 
# lapply(haar.ff.cluster, delete)
# rm(haar.ff.cluster)
# unlink("haar.ff.cluster.out.RData")
# 
# ### Try to prevent problems in R CMD check
# ## Sys.sleep(2)
# 
# 
# ### Delete all png files and temp dir
# setwd(originalDir)
# ## Sys.sleep(2)
# unlink("ADaCGH2_plot_tmp_dir", recursive = TRUE)
# ## Sys.sleep(2)
# 
# ## End(Not run)

### PNGs are in this directory
getwd()