computeMBPCR: Estimate the copy number profile

In paolarancoita/mBPCR: Bayesian Piecewise Constant Regression for DNA copy number estimation

Description

Function to estimate the copy number profile with a piecewise constant function using mBPCR. Eventually, it is possible to estimate the profile with a smoothing curve using either the Bayesian Regression Curve with K_2 (BRC with K_2) or the Bayesian Regression Curve Averaging over k (BRCAk). It is also possible to choose the estimator of the variance of the levels rhoSquare (i.e. either \hat{ρ}_1^2 or \hat{ρ}^2) and by default \hat{ρ}_1^2 is used.

Usage

 computeMBPCR(y, kMax=50, nu=NULL, rhoSquare=NULL, sigmaSquare=NULL, typeEstRho=1,
              regr=NULL)

Arguments

y	array containing the log2ratio of the copy number data
kMax	maximum number of segments
nu	mean of the segment levels. If nu=NULL, then the algorithm estimates it on the sample.
rhoSquare	variance of the segment levels. If rhoSquare=NULL, then the algorithm estimates it on the sample.
sigmaSquare	variance of the noise. If sigmaSquare=NULL, then the algorithm estimates it on the sample.
typeEstRho	choice of the estimator of rhoSquare. If typeEstRho=1, then the algorithm estimates rhoSquare with \hat{ρ}_1^2, while if typeEstRho=0, it estimates rhoSquare with \hat{ρ}^2.
regr	choice of the computation of the regression curve. If regr=NULL, then the regression curve is not computed, if regr="BRC" the Bayesian Regression Curve with K_2 is computed (BRC with K_2), if regr="BRCAk" the Bayesian Regression Curve Averaging over k is computed (BRCAk).

Details

By default, the function estimates the copy number profile with mBPCR and estimating rhoSquare on the sample, using \hat{ρ}_1^2. It is also possible to use \hat{ρ}^2 as estimator of rhoSquare, by setting typeEstRho=0, or to directly set the value of the parameter.

The function gives also the possibility to estimate the profile with a Bayesian regression curve: if regr="BRC" the Bayesian Regression Curve with K_2 is computed (BRC with K_2), if regr="BRCAk" the Bayesian Regression Curve Averaging over k is computed (BRCAk).

Value

A list containing:

estK	the estimated number of segments
estBoundaries	the estimated boundaries
estPC	the estimated profile with mBPCR
regrCurve	the estimated bayesian regression curve. It is returned only if regr!=NULL.
nu
rhoSquare
sigmaSquare
postProbT	for each probe, the posterior probablity to be a breakpoint

References

Rancoita, P. M. V., Hutter, M., Bertoni, F., Kwee, I. (2009). Bayesian DNA copy number analysis. BMC Bioinformatics 10: 10. http://www.idsia.ch/~paola/mBPCR

See Also

estProfileWithMBPCR, plotEstProfile, writeEstProfile, estGlobParam

Examples

##import the 250K NSP data of chromosome 11 of cell line JEKO-1 
data(jekoChr11Array250Knsp)


##first example 
## we select a part of chromosome 11
y <- jekoChr11Array250Knsp$log2ratio[6400:6900]
p <- jekoChr11Array250Knsp$PhysicalPosition[6400:6900]
##we estimate the profile using the global parameters estimated on the whole genome
##the profile is estimated with mBPCR and with the Bayesian Regression Curve
results <- computeMBPCR(y, nu=-3.012772e-10, rhoSquare=0.0479, sigmaSquare=0.0699, regr="BRC")
plot(p, y)
points(p, results$estPC, type='l', col='red')
points(p, results$regrCurve,type='l', col='green')

###second example 
### we select a part of chromosome 11
#y <- jekoChr11Array250Knsp$log2ratio[10600:11600]
#p <- jekoChr11Array250Knsp$PhysicalPosition[10600:11600]
###we estimate the profile using the global parameters estimated on the whole genome
###the profile is estimated with mBPCR and with the Bayesian Regression Curve Ak
#results <- computeMBPCR(y, nu=-3.012772e-10, rhoSquare=0.0479, sigmaSquare=0.0699, regr="BRCAk")
#plot(p,y)
#points(p, results$estPC, type='l', col='red')
#points(p, results$regrCurve, type='l', col='green')

paolarancoita/mBPCR documentation built on Jan. 12, 2020, 6:24 p.m.