rankednorm: Dasen Quantile Normalization by storing ranks

In bigmelon: Illumina methylation array analysis for large experiments

Description

This performs an experimental variant of dasen normalisation for .gds format objects which stores the ranks of the methylated and unmethylated intensities inside of the normalised values and interpolates the quantiles when they are needed.

Notably this eliminates a secondary re-sorting pass which is required by quantile normalisatoin as it will be performed downstream using computebeta.gds which will produce normalise betas or manually with '[' which will access the ranks and interpolate the specific quantiles as needed.

Usage

dasenrank(gds, mns, uns, onetwo, roco, calcbeta = NULL, perc = 1)
computebeta.gds(gds, new.node, mns, uns, fudge)

Arguments

gds	gds.class object which contains methylated and unmethylated intensities. The function will write two (four) nodes to this object called 'mnsrank' and 'unsrank' which contain the ranks of the given nodes.
mns	gdsn.class object OR character string that refers to location in gds that relates to the (raw) methylated intensities.
uns	gdsn.class object OR character string that refers to location in gds that relates to the (raw) unmethylated intensities.
onetwo	gdsn.class object OR character string that refers to location in gds that contains information relating to probe design OR vector of length equal to the number of rows in the array that contains 'I' and 'II' in accordance to Illumina HumanMethylation micro-array design.
roco	Sentrix (R0#C0#) position of all samples.
calcbeta	Default = NULL, if supplied with a string, a new gdsn.node will be made with supplied string, which will contain the calculated betas.
perc	A number between 0 and 1 that relates to the given proportion of columns that are used to normalise the data. Default is set to 1, but incase there are lots of samples to normalise this number can be reduce to increase speed of code.
new.node	Character string depicting name of new betas node in given gds object.
fudge	Arbitrary value to offset low intensities

Details

calcbeta is a known bottle-neck for this code! Also function is highly experimental.

Value

Nothing is returned to the R environment, however the supplied gds will have 4 or 5 gdsn.nodes added. These are: 'mnsrank', 'unsrank', 'isnamnsrank' (hidden), 'isnaunsrank'(hidden) and calcbeta if supplied. 'mnsrank' and 'unsrank' have been given some attributes - which contain the calculated quantiles from getquantilesandranks.

Author(s)

Tyler Gorrie-Stone Who to contact: <[e-mail]>

Examples

data(melon)
e <- es2gds(melon, "melon.gds")
#dasenrank(gds = e)
closefn.gds(e)
unlink("melon.gds")

bigmelon documentation built on Nov. 8, 2020, 7:40 p.m.