Rolexa (version 1.28.0)

DeCorrelateChannels: Correct for global correlations and biases

Description

Functions to correct for global correlations between color channels or between successive sequencing cycles

Usage

"DeCorrelateChannels"(int,cycles=seq(1,dim(int)[3],by=1),theta=matrix(rep(c(0.8806742,1.3727418,0.883865,1.545728),length(cycles)),ncol=4,byrow=TRUE))
"DeCorrelateChannels"(int,cycles=seq(1,dim(int)[3],by=1),theta=matrix(rep(c(0.8806742,1.3727418,0.883865,1.545728),length(cycles)),ncol=4,byrow=TRUE))
DeCorrelateChannels(int,...)
"OptimizeAngle"(int,cycles=seq(1,dim(int)[3],by=1),...)
OptimizeAngle(int,...)
"DeCorrelateCycles"(int,ncycles=dim(int)[3],rate=1.8e-2)
"DeCorrelateCycles"(int,ncycles=dim(int)[3],rate=1.8e-2)
DeCorrelateCycles(int,...)
"OptimizeRate"(int,ncycles=dim(int)[3],...)
OptimizeRate(int,...)
"TileNormalize"(run=Rolexa.env,int,cycles=seq(1,dim(int)[3],by=1))
TileNormalize(run,...)

Arguments

run

a RolexaRun object defining the run parameters

int

a SolexaIntensity object or an array

cycles, ncycles

the cycles or the number of cycles (starting from 1) to apply the correction to

theta

a length(cycles)*4 matrix with four angles per cycle defining the coordinate changes

rate

the rate of nucleotide mis-incorporation at each cycle

...

additional arguments passed to optim

Value

, DeCorrelateChannels and DeCorrelateCycles return an object of the same type as int corrected for spurious correlations. OptimizeAngle returns an length(cycles)*4 matrix and OptimizeRate returns a single positive real number.

Details

DeCorrelateChannels applies to coordinate transforms: one transforming the axes 1,2 to the axes with angles theta[,1:2] relative to axis 1, and similarly with axes 3,4 and angles theta[,3:4]. These angles can be calculated with OptimizeAngle which minimizes the correlations between channel 1 and 2, and between channel 3 and 4, for each cycle. DeCorrelateCycles assumes that at each cycles, a fraction rate of sequences fail to incorporate any nucleotides and therefore the sequence lengths at each colony display a binomial distribution which is corrected for by taking into account the intensity measured at previous cycles. OptimizeRate calculates a rate that minimizes correlations between consecutives cycles.

TileNormalize estimates the local trend by loess fitting of the model int ~ x+y and substracts it from the intensity matrix.

References

Probabilistic base calling of Solexa sequencing data, BMC Bioinformatics 2008, 9:431

See Also

TileNormalize

Examples

path = SolexaPath(system.file("extdata", package="ShortRead"))
rolenv = SetModel(idsep="_")
int = readIntensities(path,pattern="s_1_0001",withVariability=FALSE)

int1 = DeCorrelateChannels(int=int,cycles=1:5,theta=OptimizeAngle(int=int,cycles=1:5))
int2 = DeCorrelateCycles(int=int1,ncycles=5,rate=OptimizeRate(int=int1))
int3 = TileNormalize(run=rolenv,int=int,cycles=1)
seq = CombineReads(run=rolenv,path=path,pattern="s_1_0001_seq*")
PlotCycles(run=rolenv,int=int3,seq=seq,cycles=1:4)