RobLoxBioC (version 0.9)

robloxbioc: Generic Function for Preprocessing Biological Data

Description

Generic function for preprocessing biological data using optimally robust (rmx) estimators; confer Rieder (1994), Kohl (2005), Rieder et al (2008).

Usage

robloxbioc(x, ...)
# S4 method for matrix
robloxbioc(x, eps = NULL, eps.lower = 0, eps.upper = 0.05, steps = 3L, 
           fsCor = TRUE, mad0 = 1e-4)
# S4 method for AffyBatch
robloxbioc(x, bg.correct = TRUE, pmcorrect = TRUE, normalize = FALSE, 
           add.constant = 32, verbose = TRUE, eps = NULL, 
           eps.lower = 0, eps.upper = 0.05, steps = 3L, fsCor = TRUE, 
           mad0 = 1e-4, contrast.tau = 0.03, scale.tau = 10, 
           delta = 2^(-20), sc = 500)
# S4 method for beadLevelData
robloxbioc(x, channelList = list(greenChannel), probeIDs = NULL, 
           useSampleFac = FALSE, sampleFac = NULL, weightNames = "wts", 
           removeUnMappedProbes = TRUE, eps = NULL, eps.lower = 0, 
           eps.upper = 0.05, steps = 3L, fsCor = TRUE, mad0 = 1e-4)

Arguments

x

biological data.

…

additional parameters.

eps

positive real (0 < eps <= 0.5): amount of gross errors. See details below.

eps.lower

positive real (0 <= eps.lower <= eps.upper): lower bound for the amount of gross errors. See details below.

eps.upper

positive real (eps.lower <= eps.upper <= 0.5): upper bound for the amount of gross errors. See details below.

steps

positive integer. k-step is used to compute the optimally robust estimator.

fsCor

logical: perform finite-sample correction. See function finiteSampleCorrection.

mad0

scale estimate used if computed MAD is equal to zero

bg.correct

if TRUE MAS 5.0 background correction is performed; confer bg.correct.mas.

pmcorrect

method used for PM correction; TRUE calls an algorithm which is comparable to the algorithm of MAS 5.0; confer pmcorrect.mas. If FALSE only the PM intensities are used.

normalize

logical: if TRUE, Affymetrix scale normalization is performed.

add.constant

constant added to the MAS 5.0 expression values before the normalization step. Improves the variance of the measure one no longer devides by numbers close to 0 when computing fold-changes.

verbose

logical: if TRUE, some messages are printed.

contrast.tau

a number denoting the contrast tau parameter; confer the MAS 5.0 PM correction algorithm.

scale.tau

a number denoting the scale tau parameter; confer the MAS 5.0 PM correction algorithm.

delta

a number denoting the delta parameter; confer the MAS 5.0 PM correction algorithm.

sc

value at which all arrays will be scaled to.

channelList

List of objects of class illuminaChannel that defines the summarisation to be performed where in our setup only the slots transFun and name have an effect on the computations. Setting the slots outlierFun, exprFun, and varFun has no effect. In any case rmx estimators are applied.

probeIDs

Vector of ArrayAddressIDs to be included in the summarized object. The default is to summarize all probes.

useSampleFac

if TRUE sections belonging to the same biological sample will be combined. The default is to summarize each section separately.

sampleFac

optional character vector giving which a sample identifer for each section

weightNames

name of column in the beadLevelData to take extract weights

removeUnMappedProbes

if TRUE and annotation information is stored in the beadLevelData object, any ArrayAddressIDs that cannot be mapped to ILMN IDs will be removed.

Value

Return value depends on the class of x. In case of "matrix" a matrix with columns "mean" and "sd" is returned. In case of "AffyBatch" an object of class "ExpressionSet" is returned. In case of "BeadLevelData" an object of class "ExpressionSetIllumina" is returned.

Details

The optimally-robust resp. the radius-minimax (rmx) estimator for normal location and scale is used to preprocess biological data. The computation uses a k-step construction with median and MAD as starting estimators; cf. Rieder (1994) and Kohl (2005).

If the amount of gross errors (contamination) is known, it can be specified by eps. The radius of the corresponding infinitesimal contamination neighborhood (infinitesimal version of Tukey's gross error model) is obtained by multiplying eps by the square root of the sample size.

If the amount of gross errors (contamination) is unknown, which is typically the case, try to find a rough estimate for the amount of gross errors, such that it lies between eps.lower and eps.upper.

If eps is NULL, the radius-minimax (rmx) estimator in sense of Rieder et al. (2001, 2008), respectively Section 2.2 of Kohl (2005) is used.

The algorithm used for Affymetrix data is similar to MAS 5.0 (cf. Affymetrix (2002)). The main difference is the substitution of the Tukey one-step estimator by our rmx k-step (k >= 1) estimator in the PM/MM correction step. The optional scale normalization is performed as given in Affymetrix (2002).

In case of Illumina data, the rmx estimator is used to summarize the bead types. The implementation for the most part copies summarize from beadarray.

For sample size <= 2, median and MAD are used for estimation.

If eps = 0, mean and sd are computed.

References

Affymetrix, Inc. (2002). Statistical Algorithms Description Document. Affymetrix, Santa Clara.

Kohl, M. (2005) Numerical Contributions to the Asymptotic Theory of Robustness. Bayreuth: Dissertation.

Kohl M. and Deigner H.P. (2010). Preprocessing of gene expression data by optimally robust estimators. BMC Bioinformatics, 11:583.

M. Kohl, P. Ruckdeschel, and H. Rieder (2010). Infinitesimally Robust Estimation in General Smoothly Parametrized Models. Statistical Methods and Application, 19(3):333-354.

Rieder, H. (1994) Robust Asymptotic Statistics. New York: Springer.

Rieder, H., Kohl, M. and Ruckdeschel, P. (2008) The Costs of not Knowing the Radius. Statistical Methods and Applications 17(1) 13-40. Extended version: http://www.stamats.de/RRlong.pdf

See Also

roblox, rowRoblox, AffyBatch-class, generateExprVal.method.mas, ExpressionSet-class, summarize

Examples

# NOT RUN {
set.seed(123) # to have reproducible results for package checking

## similar to rowRoblox of package RobLox
ind <- rbinom(200, size=1, prob=0.05)
X <- matrix(rnorm(200, mean=ind*3, sd=(1-ind) + ind*9), nrow = 2)
robloxbioc(X)
robloxbioc(X, steps = 5)
robloxbioc(X, eps = 0.05)
robloxbioc(X, eps = 0.05, steps = 5)

## Don't run to reduce check time on CRAN
# }
# NOT RUN {
## the function is designed for large scale problems
X <- matrix(rnorm(50000*20, mean = 1), nrow = 50000)
system.time(robloxbioc(X))

## using Affymetrix-Data
## confer example to generateExprVal.method.mas
## A more worked out example can be found in the scripts folder
## of the package.
data(SpikeIn)
probes <- pm(SpikeIn) 
mas <- generateExprVal.method.mas(probes)
rl <- 2^robloxbioc(log2(t(probes)))
concentrations <- as.numeric(colnames(SpikeIn))
plot(concentrations, mas$exprs, log="xy", ylim=c(50,10000), type="b",
     ylab = "expression measures")
points(concentrations, rl[,1], pch = 20, col="orange", type="b")
legend("topleft", c("MAS", "roblox"), pch = c(1, 20))

require(affydata)
data(Dilution)
eset <- robloxbioc(Dilution)
## Affymetrix scale normalization
eset1 <- robloxbioc(Dilution, normalize = TRUE)
## "Not run" just because of computation time
require(beadarrayExampleData)
data(exampleBLData)
res <- robloxbioc(exampleBLData, eps.upper = 0.5)
res
# }