mBPCR (version 1.26.0)

computeMBPCR: Estimate the copy number profile

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 $\hat{\rho}_1^2$ or $\hat{\rho}^2$) and by default $\hat{\rho}_1^2$ 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 $\hat{\rho}_1^2$, while if typeEstRho=0, it estimates rhoSquare with $\hat{\rho}^2$.

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).

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

Details

By default, the function estimates the copy number profile with mBPCR and estimating rhoSquare on the sample, using $\hat{\rho}_1^2$. It is also possible to use $\hat{\rho}^2$ 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).

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')