readCdfUnitsWriteMap: Generates an Affymetrix cell-index write map from a CDF file

In HenrikBengtsson/affxparser: Affymetrix File Parsing SDK

Generates an Affymetrix cell-index write map from a CDF file

Description

Generates an Affymetrix cell-index write map from a CDF file.

The purpose of this method is to provide a re-ordering of cell elements such that cells in units (probesets) can be stored in contiguous blocks. When reading cell elements unit by unit, minimal file re-position is required resulting in a faster reading.

Note: At the moment does this package not provide methods to write/reorder CEL files. In the meanwhile, you have to write and re-read using your own file format. That's not too hard using writeBin() and readBin().

Usage

readCdfUnitsWriteMap(filename, units=NULL, ..., verbose=FALSE)

Arguments

filename	The pathname of the CDF file.
units	An integer vector of unit indices specifying which units to listed first. All other units are added in order at the end. If NULL, units are in order.
...	Additional arguments passed to readCdfUnits().
verbose	Either a logical, a numeric, or a Verbose object specifying how much verbose/debug information is written to standard output. If a Verbose object, how detailed the information is is specified by the threshold level of the object. If a numeric, the value is used to set the threshold of a new Verbose object. If TRUE, the threshold is set to -1 (minimal). If FALSE, no output is written (and neither is the R.utils package required).

Value

A integer vector which is a write map.

Author(s)

Henrik Bengtsson

See Also

To invert maps, see invertMap(). readCel() and readCelUnits().

Examples

  ##############################################################
if (require("AffymetrixDataTestFiles")) {            # START #
##############################################################

# Find any CDF file
cdfFile <- findCdf()

# Create a cell-index map (for writing)
writeMap <- readCdfUnitsWriteMap(cdfFile)

# Inverse map to be used to read cell elements such that, when read
# read unit by unit, they are read much faster.
readMap <- invertMap(writeMap)

# Validate the two maps
stopifnot(identical(readMap[writeMap], 1:length(readMap)))


cat("Summary of the \"randomness\" of the cell indices:\n")
moves <- diff(readMap) - 1
cat(sprintf("Number of unnecessary file re-positioning: %d (%.1f%%)\n",
                   sum(moves != 0), 100*sum(moves != 0)/length(moves)))
cat(sprintf("Extra positioning: %.1fGb\n", sum(abs(moves))/1024^3))

smallMoves <- moves[abs(moves) <= 25];
largeMoves <- moves[abs(moves)  > 25];
layout(matrix(1:2))
main <- "Non-signed file moves required in unorded file"
hist(smallMoves, nclass=51, main=main, xlab="moves <=25 bytes")
hist(largeMoves, nclass=101, main="", xlab="moves >25 bytes")

# Clean up
layout(1)
rm(cdfFile, readMap, writeMap, moves, smallMoves, largeMoves, main)

##############################################################
}                                                     # STOP #
##############################################################



  ##############################################################
if (require("AffymetrixDataTestFiles")) {            # START #
##############################################################

# - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
# Function to read Affymetrix probeset annotations
# - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
readAffymetrixProbesetAnnotation <- function(pathname, ...) {
  # Get headers
  header <- scan(pathname, what="character", sep=",", quote="\"",
                                                      quiet=TRUE, nlines=1);

  # Read only a subset of columns (unique to this example)
  cols <- c("Probe Set ID"="probeSet",
            "Chromosome"="chromosome",
            "Physical Position"="physicalPosition",
            "dbSNP RS ID"="dbSnpId");

  colClasses <- rep("NULL", length(header));
  colClasses[header %in% names(cols)] <- "character";

  # Read the data (this is what takes time)
  df <- read.table(pathname, colClasses=colClasses, header=TRUE, sep=",",
         quote="\"", na.strings="---", strip.white=TRUE, check.names=FALSE,
                 blank.lines.skip=FALSE, fill=FALSE, comment.char="", ...);

  # Re-order columns
  df <- df[,match(names(cols),colnames(df))];
  colnames(df) <- cols;

  # Use "Probe Set ID" as rownames. Note that if we use 'row.names=1'
  # or similar something goes wrong. /HB 2006-03-06
  rownames(df) <- df[[1]];
  df <- df[,-1]; 

  # Change types of columns
  df[[1]] <- factor(df[[1]], levels=c(1:22,"X","Y",NA), ordered=TRUE);
  df[[2]] <- as.integer(df[[2]]);

  df;
} # readAffymetrixProbesetAnnotation()



# - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
# Main
# - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
for (zz in 1) {
# Chip to be remapped
chipType <- "Mapping50K_Xba240"

annoFile <- paste(chipType, "_annot.csv", sep="")
cdfFile <- findCdf(chipType)
if (is.null(cdfFile) || !file.exists(annoFile))
  break;

# Read SNP location details
snpInfo <- readAffymetrixProbesetAnnotation(annoFile)

# Order by chromsome and then physical position
o <- order(snpInfo[[1]], snpInfo[[2]])
snpInfo <- snpInfo[o,]
rm(o)

# Read unit names in CDF file
unitNames <- readCdfUnitNames(cdfFile)

# The CDF unit indices sorted by chromsomal position
units <- match(rownames(snpInfo), unitNames)

# ...and cell indices in the same order
writeMap <- readCdfUnitsWriteMap(cdfFile, units=units)

# Inverse map to be used to write cell elements such that, if they
# later are read unit by unit, they are read in contiguous blocks.
readMap <- invertMap(writeMap)

# Clean up
rm(chipType, annoFile, cdfFile, snpInfo, unitNames, units, readMap, writeMap)

} # for (zz in 1)
##############################################################
}                                                     # STOP #
##############################################################

HenrikBengtsson/affxparser documentation built on Nov. 10, 2024, 11:33 a.m.